├── LICENSE
├── LSTM with multi variables.ipynb
├── README.md
├── auto_analyzer.py
├── auto_analyzer_rand.py
├── is_future_purchase.ipynb
├── purchacedata_base.csv
├── tokyo-weather-2003-2012.csv
└── 転移学習テスト.ipynb
/LICENSE:
--------------------------------------------------------------------------------
1 | GNU GENERAL PUBLIC LICENSE
2 | Version 3, 29 June 2007
3 |
4 | Copyright (C) 2007 Free Software Foundation, Inc.
5 | Everyone is permitted to copy and distribute verbatim copies
6 | of this license document, but changing it is not allowed.
7 |
8 | Preamble
9 |
10 | The GNU General Public License is a free, copyleft license for
11 | software and other kinds of works.
12 |
13 | The licenses for most software and other practical works are designed
14 | to take away your freedom to share and change the works. By contrast,
15 | the GNU General Public License is intended to guarantee your freedom to
16 | share and change all versions of a program--to make sure it remains free
17 | software for all its users. We, the Free Software Foundation, use the
18 | GNU General Public License for most of our software; it applies also to
19 | any other work released this way by its authors. You can apply it to
20 | your programs, too.
21 |
22 | When we speak of free software, we are referring to freedom, not
23 | price. Our General Public Licenses are designed to make sure that you
24 | have the freedom to distribute copies of free software (and charge for
25 | them if you wish), that you receive source code or can get it if you
26 | want it, that you can change the software or use pieces of it in new
27 | free programs, and that you know you can do these things.
28 |
29 | To protect your rights, we need to prevent others from denying you
30 | these rights or asking you to surrender the rights. Therefore, you have
31 | certain responsibilities if you distribute copies of the software, or if
32 | you modify it: responsibilities to respect the freedom of others.
33 |
34 | For example, if you distribute copies of such a program, whether
35 | gratis or for a fee, you must pass on to the recipients the same
36 | freedoms that you received. You must make sure that they, too, receive
37 | or can get the source code. And you must show them these terms so they
38 | know their rights.
39 |
40 | Developers that use the GNU GPL protect your rights with two steps:
41 | (1) assert copyright on the software, and (2) offer you this License
42 | giving you legal permission to copy, distribute and/or modify it.
43 |
44 | For the developers' and authors' protection, the GPL clearly explains
45 | that there is no warranty for this free software. For both users' and
46 | authors' sake, the GPL requires that modified versions be marked as
47 | changed, so that their problems will not be attributed erroneously to
48 | authors of previous versions.
49 |
50 | Some devices are designed to deny users access to install or run
51 | modified versions of the software inside them, although the manufacturer
52 | can do so. This is fundamentally incompatible with the aim of
53 | protecting users' freedom to change the software. The systematic
54 | pattern of such abuse occurs in the area of products for individuals to
55 | use, which is precisely where it is most unacceptable. Therefore, we
56 | have designed this version of the GPL to prohibit the practice for those
57 | products. If such problems arise substantially in other domains, we
58 | stand ready to extend this provision to those domains in future versions
59 | of the GPL, as needed to protect the freedom of users.
60 |
61 | Finally, every program is threatened constantly by software patents.
62 | States should not allow patents to restrict development and use of
63 | software on general-purpose computers, but in those that do, we wish to
64 | avoid the special danger that patents applied to a free program could
65 | make it effectively proprietary. To prevent this, the GPL assures that
66 | patents cannot be used to render the program non-free.
67 |
68 | The precise terms and conditions for copying, distribution and
69 | modification follow.
70 |
71 | TERMS AND CONDITIONS
72 |
73 | 0. Definitions.
74 |
75 | "This License" refers to version 3 of the GNU General Public License.
76 |
77 | "Copyright" also means copyright-like laws that apply to other kinds of
78 | works, such as semiconductor masks.
79 |
80 | "The Program" refers to any copyrightable work licensed under this
81 | License. Each licensee is addressed as "you". "Licensees" and
82 | "recipients" may be individuals or organizations.
83 |
84 | To "modify" a work means to copy from or adapt all or part of the work
85 | in a fashion requiring copyright permission, other than the making of an
86 | exact copy. The resulting work is called a "modified version" of the
87 | earlier work or a work "based on" the earlier work.
88 |
89 | A "covered work" means either the unmodified Program or a work based
90 | on the Program.
91 |
92 | To "propagate" a work means to do anything with it that, without
93 | permission, would make you directly or secondarily liable for
94 | infringement under applicable copyright law, except executing it on a
95 | computer or modifying a private copy. Propagation includes copying,
96 | distribution (with or without modification), making available to the
97 | public, and in some countries other activities as well.
98 |
99 | To "convey" a work means any kind of propagation that enables other
100 | parties to make or receive copies. Mere interaction with a user through
101 | a computer network, with no transfer of a copy, is not conveying.
102 |
103 | An interactive user interface displays "Appropriate Legal Notices"
104 | to the extent that it includes a convenient and prominently visible
105 | feature that (1) displays an appropriate copyright notice, and (2)
106 | tells the user that there is no warranty for the work (except to the
107 | extent that warranties are provided), that licensees may convey the
108 | work under this License, and how to view a copy of this License. If
109 | the interface presents a list of user commands or options, such as a
110 | menu, a prominent item in the list meets this criterion.
111 |
112 | 1. Source Code.
113 |
114 | The "source code" for a work means the preferred form of the work
115 | for making modifications to it. "Object code" means any non-source
116 | form of a work.
117 |
118 | A "Standard Interface" means an interface that either is an official
119 | standard defined by a recognized standards body, or, in the case of
120 | interfaces specified for a particular programming language, one that
121 | is widely used among developers working in that language.
122 |
123 | The "System Libraries" of an executable work include anything, other
124 | than the work as a whole, that (a) is included in the normal form of
125 | packaging a Major Component, but which is not part of that Major
126 | Component, and (b) serves only to enable use of the work with that
127 | Major Component, or to implement a Standard Interface for which an
128 | implementation is available to the public in source code form. A
129 | "Major Component", in this context, means a major essential component
130 | (kernel, window system, and so on) of the specific operating system
131 | (if any) on which the executable work runs, or a compiler used to
132 | produce the work, or an object code interpreter used to run it.
133 |
134 | The "Corresponding Source" for a work in object code form means all
135 | the source code needed to generate, install, and (for an executable
136 | work) run the object code and to modify the work, including scripts to
137 | control those activities. However, it does not include the work's
138 | System Libraries, or general-purpose tools or generally available free
139 | programs which are used unmodified in performing those activities but
140 | which are not part of the work. For example, Corresponding Source
141 | includes interface definition files associated with source files for
142 | the work, and the source code for shared libraries and dynamically
143 | linked subprograms that the work is specifically designed to require,
144 | such as by intimate data communication or control flow between those
145 | subprograms and other parts of the work.
146 |
147 | The Corresponding Source need not include anything that users
148 | can regenerate automatically from other parts of the Corresponding
149 | Source.
150 |
151 | The Corresponding Source for a work in source code form is that
152 | same work.
153 |
154 | 2. Basic Permissions.
155 |
156 | All rights granted under this License are granted for the term of
157 | copyright on the Program, and are irrevocable provided the stated
158 | conditions are met. This License explicitly affirms your unlimited
159 | permission to run the unmodified Program. The output from running a
160 | covered work is covered by this License only if the output, given its
161 | content, constitutes a covered work. This License acknowledges your
162 | rights of fair use or other equivalent, as provided by copyright law.
163 |
164 | You may make, run and propagate covered works that you do not
165 | convey, without conditions so long as your license otherwise remains
166 | in force. You may convey covered works to others for the sole purpose
167 | of having them make modifications exclusively for you, or provide you
168 | with facilities for running those works, provided that you comply with
169 | the terms of this License in conveying all material for which you do
170 | not control copyright. Those thus making or running the covered works
171 | for you must do so exclusively on your behalf, under your direction
172 | and control, on terms that prohibit them from making any copies of
173 | your copyrighted material outside their relationship with you.
174 |
175 | Conveying under any other circumstances is permitted solely under
176 | the conditions stated below. Sublicensing is not allowed; section 10
177 | makes it unnecessary.
178 |
179 | 3. Protecting Users' Legal Rights From Anti-Circumvention Law.
180 |
181 | No covered work shall be deemed part of an effective technological
182 | measure under any applicable law fulfilling obligations under article
183 | 11 of the WIPO copyright treaty adopted on 20 December 1996, or
184 | similar laws prohibiting or restricting circumvention of such
185 | measures.
186 |
187 | When you convey a covered work, you waive any legal power to forbid
188 | circumvention of technological measures to the extent such circumvention
189 | is effected by exercising rights under this License with respect to
190 | the covered work, and you disclaim any intention to limit operation or
191 | modification of the work as a means of enforcing, against the work's
192 | users, your or third parties' legal rights to forbid circumvention of
193 | technological measures.
194 |
195 | 4. Conveying Verbatim Copies.
196 |
197 | You may convey verbatim copies of the Program's source code as you
198 | receive it, in any medium, provided that you conspicuously and
199 | appropriately publish on each copy an appropriate copyright notice;
200 | keep intact all notices stating that this License and any
201 | non-permissive terms added in accord with section 7 apply to the code;
202 | keep intact all notices of the absence of any warranty; and give all
203 | recipients a copy of this License along with the Program.
204 |
205 | You may charge any price or no price for each copy that you convey,
206 | and you may offer support or warranty protection for a fee.
207 |
208 | 5. Conveying Modified Source Versions.
209 |
210 | You may convey a work based on the Program, or the modifications to
211 | produce it from the Program, in the form of source code under the
212 | terms of section 4, provided that you also meet all of these conditions:
213 |
214 | a) The work must carry prominent notices stating that you modified
215 | it, and giving a relevant date.
216 |
217 | b) The work must carry prominent notices stating that it is
218 | released under this License and any conditions added under section
219 | 7. This requirement modifies the requirement in section 4 to
220 | "keep intact all notices".
221 |
222 | c) You must license the entire work, as a whole, under this
223 | License to anyone who comes into possession of a copy. This
224 | License will therefore apply, along with any applicable section 7
225 | additional terms, to the whole of the work, and all its parts,
226 | regardless of how they are packaged. This License gives no
227 | permission to license the work in any other way, but it does not
228 | invalidate such permission if you have separately received it.
229 |
230 | d) If the work has interactive user interfaces, each must display
231 | Appropriate Legal Notices; however, if the Program has interactive
232 | interfaces that do not display Appropriate Legal Notices, your
233 | work need not make them do so.
234 |
235 | A compilation of a covered work with other separate and independent
236 | works, which are not by their nature extensions of the covered work,
237 | and which are not combined with it such as to form a larger program,
238 | in or on a volume of a storage or distribution medium, is called an
239 | "aggregate" if the compilation and its resulting copyright are not
240 | used to limit the access or legal rights of the compilation's users
241 | beyond what the individual works permit. Inclusion of a covered work
242 | in an aggregate does not cause this License to apply to the other
243 | parts of the aggregate.
244 |
245 | 6. Conveying Non-Source Forms.
246 |
247 | You may convey a covered work in object code form under the terms
248 | of sections 4 and 5, provided that you also convey the
249 | machine-readable Corresponding Source under the terms of this License,
250 | in one of these ways:
251 |
252 | a) Convey the object code in, or embodied in, a physical product
253 | (including a physical distribution medium), accompanied by the
254 | Corresponding Source fixed on a durable physical medium
255 | customarily used for software interchange.
256 |
257 | b) Convey the object code in, or embodied in, a physical product
258 | (including a physical distribution medium), accompanied by a
259 | written offer, valid for at least three years and valid for as
260 | long as you offer spare parts or customer support for that product
261 | model, to give anyone who possesses the object code either (1) a
262 | copy of the Corresponding Source for all the software in the
263 | product that is covered by this License, on a durable physical
264 | medium customarily used for software interchange, for a price no
265 | more than your reasonable cost of physically performing this
266 | conveying of source, or (2) access to copy the
267 | Corresponding Source from a network server at no charge.
268 |
269 | c) Convey individual copies of the object code with a copy of the
270 | written offer to provide the Corresponding Source. This
271 | alternative is allowed only occasionally and noncommercially, and
272 | only if you received the object code with such an offer, in accord
273 | with subsection 6b.
274 |
275 | d) Convey the object code by offering access from a designated
276 | place (gratis or for a charge), and offer equivalent access to the
277 | Corresponding Source in the same way through the same place at no
278 | further charge. You need not require recipients to copy the
279 | Corresponding Source along with the object code. If the place to
280 | copy the object code is a network server, the Corresponding Source
281 | may be on a different server (operated by you or a third party)
282 | that supports equivalent copying facilities, provided you maintain
283 | clear directions next to the object code saying where to find the
284 | Corresponding Source. Regardless of what server hosts the
285 | Corresponding Source, you remain obligated to ensure that it is
286 | available for as long as needed to satisfy these requirements.
287 |
288 | e) Convey the object code using peer-to-peer transmission, provided
289 | you inform other peers where the object code and Corresponding
290 | Source of the work are being offered to the general public at no
291 | charge under subsection 6d.
292 |
293 | A separable portion of the object code, whose source code is excluded
294 | from the Corresponding Source as a System Library, need not be
295 | included in conveying the object code work.
296 |
297 | A "User Product" is either (1) a "consumer product", which means any
298 | tangible personal property which is normally used for personal, family,
299 | or household purposes, or (2) anything designed or sold for incorporation
300 | into a dwelling. In determining whether a product is a consumer product,
301 | doubtful cases shall be resolved in favor of coverage. For a particular
302 | product received by a particular user, "normally used" refers to a
303 | typical or common use of that class of product, regardless of the status
304 | of the particular user or of the way in which the particular user
305 | actually uses, or expects or is expected to use, the product. A product
306 | is a consumer product regardless of whether the product has substantial
307 | commercial, industrial or non-consumer uses, unless such uses represent
308 | the only significant mode of use of the product.
309 |
310 | "Installation Information" for a User Product means any methods,
311 | procedures, authorization keys, or other information required to install
312 | and execute modified versions of a covered work in that User Product from
313 | a modified version of its Corresponding Source. The information must
314 | suffice to ensure that the continued functioning of the modified object
315 | code is in no case prevented or interfered with solely because
316 | modification has been made.
317 |
318 | If you convey an object code work under this section in, or with, or
319 | specifically for use in, a User Product, and the conveying occurs as
320 | part of a transaction in which the right of possession and use of the
321 | User Product is transferred to the recipient in perpetuity or for a
322 | fixed term (regardless of how the transaction is characterized), the
323 | Corresponding Source conveyed under this section must be accompanied
324 | by the Installation Information. But this requirement does not apply
325 | if neither you nor any third party retains the ability to install
326 | modified object code on the User Product (for example, the work has
327 | been installed in ROM).
328 |
329 | The requirement to provide Installation Information does not include a
330 | requirement to continue to provide support service, warranty, or updates
331 | for a work that has been modified or installed by the recipient, or for
332 | the User Product in which it has been modified or installed. Access to a
333 | network may be denied when the modification itself materially and
334 | adversely affects the operation of the network or violates the rules and
335 | protocols for communication across the network.
336 |
337 | Corresponding Source conveyed, and Installation Information provided,
338 | in accord with this section must be in a format that is publicly
339 | documented (and with an implementation available to the public in
340 | source code form), and must require no special password or key for
341 | unpacking, reading or copying.
342 |
343 | 7. Additional Terms.
344 |
345 | "Additional permissions" are terms that supplement the terms of this
346 | License by making exceptions from one or more of its conditions.
347 | Additional permissions that are applicable to the entire Program shall
348 | be treated as though they were included in this License, to the extent
349 | that they are valid under applicable law. If additional permissions
350 | apply only to part of the Program, that part may be used separately
351 | under those permissions, but the entire Program remains governed by
352 | this License without regard to the additional permissions.
353 |
354 | When you convey a copy of a covered work, you may at your option
355 | remove any additional permissions from that copy, or from any part of
356 | it. (Additional permissions may be written to require their own
357 | removal in certain cases when you modify the work.) You may place
358 | additional permissions on material, added by you to a covered work,
359 | for which you have or can give appropriate copyright permission.
360 |
361 | Notwithstanding any other provision of this License, for material you
362 | add to a covered work, you may (if authorized by the copyright holders of
363 | that material) supplement the terms of this License with terms:
364 |
365 | a) Disclaiming warranty or limiting liability differently from the
366 | terms of sections 15 and 16 of this License; or
367 |
368 | b) Requiring preservation of specified reasonable legal notices or
369 | author attributions in that material or in the Appropriate Legal
370 | Notices displayed by works containing it; or
371 |
372 | c) Prohibiting misrepresentation of the origin of that material, or
373 | requiring that modified versions of such material be marked in
374 | reasonable ways as different from the original version; or
375 |
376 | d) Limiting the use for publicity purposes of names of licensors or
377 | authors of the material; or
378 |
379 | e) Declining to grant rights under trademark law for use of some
380 | trade names, trademarks, or service marks; or
381 |
382 | f) Requiring indemnification of licensors and authors of that
383 | material by anyone who conveys the material (or modified versions of
384 | it) with contractual assumptions of liability to the recipient, for
385 | any liability that these contractual assumptions directly impose on
386 | those licensors and authors.
387 |
388 | All other non-permissive additional terms are considered "further
389 | restrictions" within the meaning of section 10. If the Program as you
390 | received it, or any part of it, contains a notice stating that it is
391 | governed by this License along with a term that is a further
392 | restriction, you may remove that term. If a license document contains
393 | a further restriction but permits relicensing or conveying under this
394 | License, you may add to a covered work material governed by the terms
395 | of that license document, provided that the further restriction does
396 | not survive such relicensing or conveying.
397 |
398 | If you add terms to a covered work in accord with this section, you
399 | must place, in the relevant source files, a statement of the
400 | additional terms that apply to those files, or a notice indicating
401 | where to find the applicable terms.
402 |
403 | Additional terms, permissive or non-permissive, may be stated in the
404 | form of a separately written license, or stated as exceptions;
405 | the above requirements apply either way.
406 |
407 | 8. Termination.
408 |
409 | You may not propagate or modify a covered work except as expressly
410 | provided under this License. Any attempt otherwise to propagate or
411 | modify it is void, and will automatically terminate your rights under
412 | this License (including any patent licenses granted under the third
413 | paragraph of section 11).
414 |
415 | However, if you cease all violation of this License, then your
416 | license from a particular copyright holder is reinstated (a)
417 | provisionally, unless and until the copyright holder explicitly and
418 | finally terminates your license, and (b) permanently, if the copyright
419 | holder fails to notify you of the violation by some reasonable means
420 | prior to 60 days after the cessation.
421 |
422 | Moreover, your license from a particular copyright holder is
423 | reinstated permanently if the copyright holder notifies you of the
424 | violation by some reasonable means, this is the first time you have
425 | received notice of violation of this License (for any work) from that
426 | copyright holder, and you cure the violation prior to 30 days after
427 | your receipt of the notice.
428 |
429 | Termination of your rights under this section does not terminate the
430 | licenses of parties who have received copies or rights from you under
431 | this License. If your rights have been terminated and not permanently
432 | reinstated, you do not qualify to receive new licenses for the same
433 | material under section 10.
434 |
435 | 9. Acceptance Not Required for Having Copies.
436 |
437 | You are not required to accept this License in order to receive or
438 | run a copy of the Program. Ancillary propagation of a covered work
439 | occurring solely as a consequence of using peer-to-peer transmission
440 | to receive a copy likewise does not require acceptance. However,
441 | nothing other than this License grants you permission to propagate or
442 | modify any covered work. These actions infringe copyright if you do
443 | not accept this License. Therefore, by modifying or propagating a
444 | covered work, you indicate your acceptance of this License to do so.
445 |
446 | 10. Automatic Licensing of Downstream Recipients.
447 |
448 | Each time you convey a covered work, the recipient automatically
449 | receives a license from the original licensors, to run, modify and
450 | propagate that work, subject to this License. You are not responsible
451 | for enforcing compliance by third parties with this License.
452 |
453 | An "entity transaction" is a transaction transferring control of an
454 | organization, or substantially all assets of one, or subdividing an
455 | organization, or merging organizations. If propagation of a covered
456 | work results from an entity transaction, each party to that
457 | transaction who receives a copy of the work also receives whatever
458 | licenses to the work the party's predecessor in interest had or could
459 | give under the previous paragraph, plus a right to possession of the
460 | Corresponding Source of the work from the predecessor in interest, if
461 | the predecessor has it or can get it with reasonable efforts.
462 |
463 | You may not impose any further restrictions on the exercise of the
464 | rights granted or affirmed under this License. For example, you may
465 | not impose a license fee, royalty, or other charge for exercise of
466 | rights granted under this License, and you may not initiate litigation
467 | (including a cross-claim or counterclaim in a lawsuit) alleging that
468 | any patent claim is infringed by making, using, selling, offering for
469 | sale, or importing the Program or any portion of it.
470 |
471 | 11. Patents.
472 |
473 | A "contributor" is a copyright holder who authorizes use under this
474 | License of the Program or a work on which the Program is based. The
475 | work thus licensed is called the contributor's "contributor version".
476 |
477 | A contributor's "essential patent claims" are all patent claims
478 | owned or controlled by the contributor, whether already acquired or
479 | hereafter acquired, that would be infringed by some manner, permitted
480 | by this License, of making, using, or selling its contributor version,
481 | but do not include claims that would be infringed only as a
482 | consequence of further modification of the contributor version. For
483 | purposes of this definition, "control" includes the right to grant
484 | patent sublicenses in a manner consistent with the requirements of
485 | this License.
486 |
487 | Each contributor grants you a non-exclusive, worldwide, royalty-free
488 | patent license under the contributor's essential patent claims, to
489 | make, use, sell, offer for sale, import and otherwise run, modify and
490 | propagate the contents of its contributor version.
491 |
492 | In the following three paragraphs, a "patent license" is any express
493 | agreement or commitment, however denominated, not to enforce a patent
494 | (such as an express permission to practice a patent or covenant not to
495 | sue for patent infringement). To "grant" such a patent license to a
496 | party means to make such an agreement or commitment not to enforce a
497 | patent against the party.
498 |
499 | If you convey a covered work, knowingly relying on a patent license,
500 | and the Corresponding Source of the work is not available for anyone
501 | to copy, free of charge and under the terms of this License, through a
502 | publicly available network server or other readily accessible means,
503 | then you must either (1) cause the Corresponding Source to be so
504 | available, or (2) arrange to deprive yourself of the benefit of the
505 | patent license for this particular work, or (3) arrange, in a manner
506 | consistent with the requirements of this License, to extend the patent
507 | license to downstream recipients. "Knowingly relying" means you have
508 | actual knowledge that, but for the patent license, your conveying the
509 | covered work in a country, or your recipient's use of the covered work
510 | in a country, would infringe one or more identifiable patents in that
511 | country that you have reason to believe are valid.
512 |
513 | If, pursuant to or in connection with a single transaction or
514 | arrangement, you convey, or propagate by procuring conveyance of, a
515 | covered work, and grant a patent license to some of the parties
516 | receiving the covered work authorizing them to use, propagate, modify
517 | or convey a specific copy of the covered work, then the patent license
518 | you grant is automatically extended to all recipients of the covered
519 | work and works based on it.
520 |
521 | A patent license is "discriminatory" if it does not include within
522 | the scope of its coverage, prohibits the exercise of, or is
523 | conditioned on the non-exercise of one or more of the rights that are
524 | specifically granted under this License. You may not convey a covered
525 | work if you are a party to an arrangement with a third party that is
526 | in the business of distributing software, under which you make payment
527 | to the third party based on the extent of your activity of conveying
528 | the work, and under which the third party grants, to any of the
529 | parties who would receive the covered work from you, a discriminatory
530 | patent license (a) in connection with copies of the covered work
531 | conveyed by you (or copies made from those copies), or (b) primarily
532 | for and in connection with specific products or compilations that
533 | contain the covered work, unless you entered into that arrangement,
534 | or that patent license was granted, prior to 28 March 2007.
535 |
536 | Nothing in this License shall be construed as excluding or limiting
537 | any implied license or other defenses to infringement that may
538 | otherwise be available to you under applicable patent law.
539 |
540 | 12. No Surrender of Others' Freedom.
541 |
542 | If conditions are imposed on you (whether by court order, agreement or
543 | otherwise) that contradict the conditions of this License, they do not
544 | excuse you from the conditions of this License. If you cannot convey a
545 | covered work so as to satisfy simultaneously your obligations under this
546 | License and any other pertinent obligations, then as a consequence you may
547 | not convey it at all. For example, if you agree to terms that obligate you
548 | to collect a royalty for further conveying from those to whom you convey
549 | the Program, the only way you could satisfy both those terms and this
550 | License would be to refrain entirely from conveying the Program.
551 |
552 | 13. Use with the GNU Affero General Public License.
553 |
554 | Notwithstanding any other provision of this License, you have
555 | permission to link or combine any covered work with a work licensed
556 | under version 3 of the GNU Affero General Public License into a single
557 | combined work, and to convey the resulting work. The terms of this
558 | License will continue to apply to the part which is the covered work,
559 | but the special requirements of the GNU Affero General Public License,
560 | section 13, concerning interaction through a network will apply to the
561 | combination as such.
562 |
563 | 14. Revised Versions of this License.
564 |
565 | The Free Software Foundation may publish revised and/or new versions of
566 | the GNU General Public License from time to time. Such new versions will
567 | be similar in spirit to the present version, but may differ in detail to
568 | address new problems or concerns.
569 |
570 | Each version is given a distinguishing version number. If the
571 | Program specifies that a certain numbered version of the GNU General
572 | Public License "or any later version" applies to it, you have the
573 | option of following the terms and conditions either of that numbered
574 | version or of any later version published by the Free Software
575 | Foundation. If the Program does not specify a version number of the
576 | GNU General Public License, you may choose any version ever published
577 | by the Free Software Foundation.
578 |
579 | If the Program specifies that a proxy can decide which future
580 | versions of the GNU General Public License can be used, that proxy's
581 | public statement of acceptance of a version permanently authorizes you
582 | to choose that version for the Program.
583 |
584 | Later license versions may give you additional or different
585 | permissions. However, no additional obligations are imposed on any
586 | author or copyright holder as a result of your choosing to follow a
587 | later version.
588 |
589 | 15. Disclaimer of Warranty.
590 |
591 | THERE IS NO WARRANTY FOR THE PROGRAM, TO THE EXTENT PERMITTED BY
592 | APPLICABLE LAW. EXCEPT WHEN OTHERWISE STATED IN WRITING THE COPYRIGHT
593 | HOLDERS AND/OR OTHER PARTIES PROVIDE THE PROGRAM "AS IS" WITHOUT WARRANTY
594 | OF ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING, BUT NOT LIMITED TO,
595 | THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
596 | PURPOSE. THE ENTIRE RISK AS TO THE QUALITY AND PERFORMANCE OF THE PROGRAM
597 | IS WITH YOU. SHOULD THE PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF
598 | ALL NECESSARY SERVICING, REPAIR OR CORRECTION.
599 |
600 | 16. Limitation of Liability.
601 |
602 | IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING
603 | WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MODIFIES AND/OR CONVEYS
604 | THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES, INCLUDING ANY
605 | GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING OUT OF THE
606 | USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED TO LOSS OF
607 | DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD
608 | PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER PROGRAMS),
609 | EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE POSSIBILITY OF
610 | SUCH DAMAGES.
611 |
612 | 17. Interpretation of Sections 15 and 16.
613 |
614 | If the disclaimer of warranty and limitation of liability provided
615 | above cannot be given local legal effect according to their terms,
616 | reviewing courts shall apply local law that most closely approximates
617 | an absolute waiver of all civil liability in connection with the
618 | Program, unless a warranty or assumption of liability accompanies a
619 | copy of the Program in return for a fee.
620 |
621 | END OF TERMS AND CONDITIONS
622 |
623 | How to Apply These Terms to Your New Programs
624 |
625 | If you develop a new program, and you want it to be of the greatest
626 | possible use to the public, the best way to achieve this is to make it
627 | free software which everyone can redistribute and change under these terms.
628 |
629 | To do so, attach the following notices to the program. It is safest
630 | to attach them to the start of each source file to most effectively
631 | state the exclusion of warranty; and each file should have at least
632 | the "copyright" line and a pointer to where the full notice is found.
633 |
634 | {one line to give the program's name and a brief idea of what it does.}
635 | Copyright (C) {year} {name of author}
636 |
637 | This program is free software: you can redistribute it and/or modify
638 | it under the terms of the GNU General Public License as published by
639 | the Free Software Foundation, either version 3 of the License, or
640 | (at your option) any later version.
641 |
642 | This program is distributed in the hope that it will be useful,
643 | but WITHOUT ANY WARRANTY; without even the implied warranty of
644 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
645 | GNU General Public License for more details.
646 |
647 | You should have received a copy of the GNU General Public License
648 | along with this program. If not, see .
649 |
650 | Also add information on how to contact you by electronic and paper mail.
651 |
652 | If the program does terminal interaction, make it output a short
653 | notice like this when it starts in an interactive mode:
654 |
655 | {project} Copyright (C) {year} {fullname}
656 | This program comes with ABSOLUTELY NO WARRANTY; for details type `show w'.
657 | This is free software, and you are welcome to redistribute it
658 | under certain conditions; type `show c' for details.
659 |
660 | The hypothetical commands `show w' and `show c' should show the appropriate
661 | parts of the General Public License. Of course, your program's commands
662 | might be different; for a GUI interface, you would use an "about box".
663 |
664 | You should also get your employer (if you work as a programmer) or school,
665 | if any, to sign a "copyright disclaimer" for the program, if necessary.
666 | For more information on this, and how to apply and follow the GNU GPL, see
667 | .
668 |
669 | The GNU General Public License does not permit incorporating your program
670 | into proprietary programs. If your program is a subroutine library, you
671 | may consider it more useful to permit linking proprietary applications with
672 | the library. If this is what you want to do, use the GNU Lesser General
673 | Public License instead of this License. But first, please read
674 | .
675 |
--------------------------------------------------------------------------------
/README.md:
--------------------------------------------------------------------------------
1 | # auto_analyzer.py
2 | - コマンドラインでデータセットのファイル名などと共に呼び出すことで、自動で分析を始めて完成したモデルを保存します。
3 | - 保存したモデルを使って予測を行い、結果を別のファイルに出力します。
4 | - 自動でバッチサイズ、エポック、ニューロン数、レイヤー数、ラーニングレートを変えてチューニングします。
5 |
6 | ## auto_analyzer_rand.py
7 | パラメタの組み合わせを総当りでなくランダムで学習します。学習に時間がかかりすぎる場合にお試しください。
8 |
9 | ## 準備するもの
10 | - 1行目にヘッダーを付けたデータセット
11 |
12 | ・ CSV形式を推奨
13 |
14 | ・ 欠損値がある行は自動で除外されます
15 |
16 | ・ 外れ値は修正しておいてください
17 |
18 | ・ 予測データのカテゴリ値の種類は、学習時と同じ数だけ入れてください
19 |
20 | - auto_analyzer.pyとデータセットを同じディレクトリに設置
21 | - 格納ディレクトリに書き込み権限を付与
22 |
23 | ## 使い方
24 | 1.コマンドラインを開く
25 |
26 | 2.データセットと本コードが格納されているディレクトリへ移動
27 |
28 | 3.「python auto_analyzer.py + パラメタ値」で呼び出す
29 |
30 | ### 呼び出し例
31 | python auto_analyzer.py --mode create --input_file xxx.csv --method regression --model_file test --definition str,int,int
32 |
33 | ## パラメタ説明
34 | ### --mode(必須)
35 | create:新しくモデルを作成する
36 |
37 | predict:作ったモデルで予測する
38 |
39 | ### --input_file(必須)
40 | データセットの入ったファイル名を入れる
41 |
42 | ### --method(必須)
43 | binary:二値分類
44 |
45 | multiple:多値分類
46 |
47 | regression:回帰
48 |
49 | ### --output_file(非必須)
50 | 予測時に結果を格納するファイル名を入れる
51 |
52 | ### --model_file(非必須)
53 | モデルを保存/読み込む際に参照
54 |
55 | ### --definition(非必須)
56 | データ型が自動認識で問題がある場合に入力してください。
57 | 全カラム分のデータ型を[str, int, float]の中から選び、「,」カンマ区切りで入れてください。
58 |
59 | ## 動作環境
60 | - python 3.5.0
61 |
62 | - keras 2.0.3
63 |
64 |
--------------------------------------------------------------------------------
/auto_analyzer.py:
--------------------------------------------------------------------------------
1 | import numpy
2 | import pandas
3 | from keras.models import Sequential
4 | from keras.layers import Dense
5 | from keras import optimizers
6 | from keras.wrappers.scikit_learn import KerasRegressor
7 | from keras.wrappers.scikit_learn import KerasClassifier
8 | from sklearn.model_selection import cross_val_score
9 | from sklearn.model_selection import KFold
10 | from sklearn.grid_search import GridSearchCV
11 | from sklearn.preprocessing import StandardScaler
12 | from sklearn.pipeline import Pipeline
13 | from keras.models import load_model
14 | import os
15 | import argparse
16 |
17 | #----------------------------
18 | # get command line variables
19 | #----------------------------
20 | parser = argparse.ArgumentParser(description='Make models by keras. Place Y on the head column in the cleaned dataset with header names on the top row. Rows containing null values will be deleated.')
21 | parser.add_argument('--mode', choices=['create', 'predict'], dest='mode', metavar='create/predict', type=str, nargs='+', required=True,
22 | help='an integer for the accumulator')
23 | parser.add_argument('--input_file', dest='input_file', type=str, nargs='+', required=True,
24 | help='path to dataset or model')
25 | parser.add_argument('--method', choices=['binary', 'multiple', 'regression'], metavar='binary/multiple/regression', dest='method', type=str, nargs='+', required=True,
26 | help='Model type you solve')
27 | parser.add_argument('--output_file', dest='output_file', default=False, required=False,
28 | help='If you input output_file it will save result as directed path.')
29 | parser.add_argument('--model_file', dest='model_file', default=False, nargs='*',
30 | help='If you input model_file it will save or load a model.')
31 | parser.add_argument('--definition', metavar='array of data type such as str, int and float with delimiter [,]', dest='definition', default=False, nargs='*',
32 | help='If you define data type of columns, send array of full column definitions.')
33 |
34 | args = parser.parse_args()
35 |
36 | #----------------------------
37 | # functions
38 | #----------------------------
39 | class MakeModel:
40 | #init
41 | def __init__(self, args):
42 | self.X = self.Y = []
43 | self.row_length = self.column_length = 0
44 | self.method = args.method[0]
45 | self.ifp = args.input_file[0]
46 |
47 | if args.model_file != False:
48 | self.mfp = args.model_file[0]
49 | else:
50 | self.mfp = False
51 |
52 | if args.output_file != False:
53 | self.ofp = args.output_file[0]
54 | else:
55 | self.ofp = False
56 |
57 | if args.definition != False:
58 | self.dfin = args.definition.split(",")
59 | else:
60 | self.dfin = False
61 |
62 | #create layers
63 | def create_model(self, evMethod, neurons, layers, act, learn_rate, cls, mtr):
64 | # Criate model
65 | model = Sequential()
66 | model.add(Dense(neurons, input_dim=self.column_length, kernel_initializer='normal', activation='relu'))
67 | for i in range(1, layers):
68 | model.add(Dense(int(numpy.ceil(numpy.power(neurons,1/i)*2)), kernel_initializer='normal', activation='relu'))
69 | model.add(Dense(cls, kernel_initializer='normal', activation=act))
70 | # Compile model
71 | adam = optimizers.Adam(lr=learn_rate)
72 | model.compile(loss=evMethod, optimizer=adam, metrics=mtr)
73 | return model
74 |
75 | #load dataset
76 | def load_dataset(self):
77 | dataframe = pandas.read_csv(self.ifp, header=0).dropna()
78 | if self.dfin != False:
79 | dataframe[dataframe.columns].apply(lambda x: x.astype(self.dfin[dataframe.columns.get_loc(x.name)]))
80 | dataframe_X = pandas.get_dummies(dataframe[dataframe.columns[1:]]) #create dummy variables
81 | if self.method == 'multiple':
82 | dataframe_Y = pandas.get_dummies(dataframe[dataframe.columns[0]]) #create dummy variables
83 | else:
84 | dataframe_Y = dataframe[dataframe.columns[0]]
85 | #print(dataframe_Y.head())
86 | #print(dataframe_X.head())
87 | self.row_length, self.column_length = dataframe_X.shape
88 | self.X = dataframe_X.values
89 | self.Y = dataframe_Y.values
90 |
91 | #train
92 | def train_model(self):
93 | #pipe to Grid Search
94 | estimators = []
95 | estimators.append(('standardize', StandardScaler()))
96 |
97 | #rely on chosen method parameters
98 | if self.method == 'binary':
99 | evMethod = ['binary_crossentropy']
100 | activation = ['sigmoid']
101 | metr = [['accuracy']]
102 | estimators.append(('mlp', KerasClassifier(build_fn=self.create_model, epochs=10, batch_size=200, verbose=1)))
103 | cls = [1]
104 | elif self.method == 'multiple':
105 | evMethod = [['categorical_crossentropy']]
106 | activation = ['softmax']
107 | metr = [['accuracy']]
108 | estimators.append(('mlp', KerasClassifier(build_fn=self.create_model, epochs=10, batch_size=200, verbose=1)))
109 | cls = [self.Y.shape[1]]
110 | else:
111 | evMethod = ['mean_squared_error']
112 | activation = [None]
113 | metr = [None]
114 | estimators.append(('mlp', KerasRegressor(build_fn=self.create_model, epochs=10, batch_size=200, verbose=1)))
115 | cls = [1]
116 |
117 | pipeline = Pipeline(estimators)
118 |
119 | #test parameters
120 | batch_size = list(set([int(numpy.ceil(self.row_length/i)) for i in [1000,300,100]]))
121 | epochs = [10, 50, 100]
122 | neurons = list(set([int(numpy.ceil(self.column_length/i)*2) for i in numpy.arange(1,3,0.4)]))
123 | learn_rate = [0.001, 0.005, 0.01, 0.07]
124 | layers = [1,2,3,4,5]
125 | #test parameter
126 | """batch_size = [31]
127 | epochs = [100]
128 | neurons = [32]
129 | learn_rate = [0.01]
130 | layers = [5]"""
131 | #execution
132 | param_grid = dict(mlp__neurons = neurons, mlp__batch_size = batch_size, mlp__epochs=epochs, mlp__learn_rate=learn_rate, mlp__layers=layers, mlp__act=activation, mlp__evMethod=evMethod, mlp__cls=cls, mlp__mtr=metr)
133 | grid = GridSearchCV(estimator=pipeline, param_grid=param_grid)
134 | grid_result = grid.fit(self.X, self.Y)
135 |
136 | #output best parameter condition
137 | clf = []
138 | clf = grid_result.best_estimator_
139 | print(clf.get_params())
140 | accuracy = clf.score(self.X, self.Y)
141 | if self.method in ['binary', 'multiple']:
142 | print("\nAccuracy: %.2f" % (accuracy))
143 | else:
144 | print("Results: %.2f (%.2f) MSE" % (accuracy.mean(), accuracy.std()))
145 |
146 | #save model
147 | if self.mfp != False:
148 | clf.steps[1][1].model.save(self.mfp)
149 |
150 | #predict dataset
151 | def predict_ds(self):
152 | model = load_model(self.mfp)
153 | model.summary()
154 | sc = StandardScaler()
155 | self.X = sc.fit_transform(self.X)
156 | pr_Y = model.predict(self.X)
157 | if len([self.Y != '__null__']) > 0:
158 | if self.method == 'binary':
159 | predictions = [float(numpy.round(x)) for x in pr_Y]
160 | accuracy = numpy.mean(predictions == self.Y)
161 | print("Prediction Accuracy: %.2f%%" % (accuracy*100))
162 | elif self.method == 'multiple':
163 | predictions = []
164 | for i in range(0, len(pr_Y)-1):
165 | for j in range(0, len(pr_Y[i])-1):
166 | predictions.append(int(round(pr_Y[i][j]) - self.Y[i][j]))
167 | accuracy_total = len([x for x in predictions if x == 0])/len(predictions)
168 | accuracy_tooneg = len([x for x in predictions if x == -1])/len(predictions)
169 | accuracy_toopos = len([x for x in predictions if x == 1])/len(predictions)
170 | print("Prediction Accuracy: %.2f%% (positive-error:%.2f%%/negative-error:%.2f%%)" % (accuracy_total*100, accuracy_tooneg*100, accuracy_toopos*100))
171 | else:
172 | accuracy = numpy.mean((self.Y - pr_Y)**2)
173 | print("MSE: %.2f" % (numpy.sqrt(accuracy)))
174 |
175 | #save predicted result
176 | if self.ofp != False:
177 | numpy.savetxt(self.ofp, pr_Y, fmt='%5s')
178 |
179 | #----------------------------
180 | # select mode
181 | #----------------------------
182 | m = MakeModel(args)
183 | if args.mode == ['create']:
184 | #make model
185 | m.load_dataset()
186 | m.train_model()
187 | else:
188 | #predict dataset
189 | m.load_dataset()
190 | m.predict_ds()
191 |
--------------------------------------------------------------------------------
/auto_analyzer_rand.py:
--------------------------------------------------------------------------------
1 | import numpy
2 | import pandas
3 | from keras.models import Sequential
4 | from keras.layers import Dense
5 | from keras import optimizers
6 | from keras.wrappers.scikit_learn import KerasRegressor
7 | from keras.wrappers.scikit_learn import KerasClassifier
8 | from sklearn.model_selection import cross_val_score
9 | from sklearn.model_selection import KFold
10 | from sklearn.model_selection import RandomizedSearchCV
11 | from sklearn.preprocessing import StandardScaler
12 | from sklearn.pipeline import Pipeline
13 | from keras.models import load_model
14 | import os
15 | import argparse
16 |
17 | #----------------------------
18 | # get command line variables
19 | #----------------------------
20 | parser = argparse.ArgumentParser(description='Make models by keras. Place Y on the head column in the cleaned dataset with header names on the top row. Rows containing null values will be deleated.')
21 | parser.add_argument('--mode', choices=['create', 'predict'], dest='mode', metavar='create/predict', type=str, nargs='+', required=True,
22 | help='an integer for the accumulator')
23 | parser.add_argument('--input_file', dest='input_file', type=str, nargs='+', required=True,
24 | help='path to dataset or model')
25 | parser.add_argument('--method', choices=['binary', 'multiple', 'regression'], metavar='binary/multiple/regression', dest='method', type=str, nargs='+', required=True,
26 | help='Model type you solve')
27 | parser.add_argument('--output_file', dest='output_file', default=False, nargs='*',
28 | help='If you input output_file it will save result as directed path.')
29 | parser.add_argument('--model_file', dest='model_file', default=False, nargs='*',
30 | help='If you input model_file it will save or load a model.')
31 | parser.add_argument('--definition', metavar='array of data type such as str, int and float with delimiter [,]', dest='definition', default=False, nargs='*',
32 | help='If you define data type of columns, send array of full column definitions.')
33 |
34 | args = parser.parse_args()
35 |
36 | #----------------------------
37 | # functions
38 | #----------------------------
39 | class MakeModel:
40 | #init
41 | def __init__(self, args):
42 | self.X = self.Y = []
43 | self.row_length = self.column_length = 0
44 | self.method = args.method[0]
45 | self.ifp = args.input_file[0]
46 |
47 | if args.model_file != False:
48 | self.mfp = args.model_file[0]
49 | else:
50 | self.mfp = False
51 |
52 | if args.output_file != False:
53 | self.ofp = args.output_file[0]
54 | else:
55 | self.ofp = False
56 |
57 | if args.definition != False:
58 | self.dfin = args.definition.split(",")
59 | else:
60 | self.dfin = False
61 |
62 | #create layers
63 | def create_model(self, evMethod, neurons, layers, act, learn_rate, cls, mtr):
64 | # Criate model
65 | model = Sequential()
66 | model.add(Dense(neurons, input_dim=self.column_length, kernel_initializer='normal', activation='relu'))
67 | for i in range(2, layers+1):
68 | model.add(Dense(int(numpy.ceil(numpy.power(neurons,1/i)*2)), kernel_initializer='normal', activation='relu'))
69 | model.add(Dense(cls, kernel_initializer='normal', activation=act))
70 | # Compile model
71 | adam = optimizers.Adam(lr=learn_rate)
72 | model.compile(loss=evMethod, optimizer=adam, metrics=mtr)
73 | return model
74 |
75 | #load dataset
76 | def load_dataset(self):
77 | dataframe = pandas.read_csv(self.ifp, header=0).dropna()
78 | if self.dfin != False:
79 | dataframe[dataframe.columns].apply(lambda x: x.astype(self.dfin[dataframe.columns.get_loc(x.name)]))
80 | dataframe_X = pandas.get_dummies(dataframe[dataframe.columns[1:]]) #create dummy variables
81 | if self.method == 'multiple':
82 | dataframe_Y = pandas.get_dummies(dataframe[dataframe.columns[0]]) #create dummy variables
83 | else:
84 | dataframe_Y = dataframe[dataframe.columns[0]]
85 | self.row_length, self.column_length = dataframe_X.shape
86 | self.X = dataframe_X.values
87 | self.Y = dataframe_Y.values
88 |
89 | #train
90 | def train_model(self):
91 | #pipe to Grid Search
92 | estimators = []
93 | estimators.append(('standardize', StandardScaler()))
94 |
95 | #rely on chosen method parameters
96 | if self.method == 'binary':
97 | evMethod = ['binary_crossentropy']
98 | activation = ['sigmoid']
99 | metr = [['accuracy']]
100 | estimators.append(('mlp', KerasClassifier(build_fn=self.create_model, epochs=10, batch_size=200, verbose=1)))
101 | cls = [1]
102 | elif self.method == 'multiple':
103 | evMethod = [['categorical_crossentropy']]
104 | activation = ['softmax']
105 | metr = [['accuracy']]
106 | estimators.append(('mlp', KerasClassifier(build_fn=self.create_model, epochs=10, batch_size=200, verbose=1)))
107 | cls = [self.Y.shape[1]]
108 | else:
109 | evMethod = ['mean_squared_error']
110 | activation = [None]
111 | metr = [None]
112 | estimators.append(('mlp', KerasRegressor(build_fn=self.create_model, epochs=10, batch_size=200, verbose=1)))
113 | cls = [1]
114 |
115 | pipeline = Pipeline(estimators)
116 |
117 | #test parameters
118 | batch_size = list(set([int(numpy.ceil(self.row_length/i)) for i in [1000,300,100]]))
119 | epochs = [10, 50, 100]
120 | neurons = list(set([int(numpy.ceil(self.column_length/i)*2) for i in numpy.arange(1,3,0.4)]))
121 | learn_rate = [0.001, 0.005, 0.01, 0.07]
122 | layers = [1,2,3,4,5]
123 | #test parameter
124 | """batch_size = [10]
125 | epochs = [100]
126 | neurons = [self.column_length]
127 | learn_rate = [0.001]
128 | layers = [1]"""
129 | #execution
130 | n_iter_search = 30
131 | param_grid = dict(mlp__neurons = neurons, mlp__batch_size = batch_size, mlp__epochs=epochs, mlp__learn_rate=learn_rate, mlp__layers=layers, mlp__act=activation, mlp__evMethod=evMethod, mlp__cls=cls, mlp__mtr=metr)
132 | grid = RandomizedSearchCV(estimator=pipeline, param_distributions=param_grid, n_iter=n_iter_search)
133 | grid_result = grid.fit(self.X, self.Y)
134 | grid_result.predict(self.X) #refit weight of each variables
135 |
136 | #output best parameter condition
137 | clf = []
138 | clf = grid_result.best_estimator_
139 | print(clf.get_params())
140 | accuracy = clf.score(self.X, self.Y)
141 | if self.method in ['binary', 'multiple']:
142 | print("\nAccuracy: %.2f" % (accuracy))
143 | else:
144 | print("Results: %.2f (%.2f) MSE" % (accuracy.mean(), accuracy.std()))
145 |
146 | #save model
147 | if self.mfp != False:
148 | clf.steps[1][1].model.save(self.mfp)
149 |
150 | #predict dataset
151 | def predict_ds(self):
152 | model = load_model(self.mfp)
153 | model.summary()
154 | sc = StandardScaler()
155 | self.X = sc.fit_transform(self.X)
156 | pr_Y = model.predict(self.X)
157 | if len([self.Y != '__null__']) > 0:
158 | if self.method == 'binary':
159 | predictions = [float(numpy.round(x)) for x in pr_Y]
160 | accuracy = numpy.mean(predictions == self.Y)
161 | print("Prediction Accuracy: %.2f%%" % (accuracy*100))
162 | elif self.method == 'multiple':
163 | predictions = []
164 | for i in range(0, len(pr_Y)-1):
165 | for j in range(0, len(pr_Y[i])-1):
166 | predictions.append(int(round(pr_Y[i][j]) - self.Y[i][j]))
167 | accuracy_total = len([x for x in predictions if x == 0])/len(predictions)
168 | accuracy_tooneg = len([x for x in predictions if x == -1])/len(predictions)
169 | accuracy_toopos = len([x for x in predictions if x == 1])/len(predictions)
170 | print("Prediction Accuracy: %.2f%% (positive-error:%.2f%%/negative-error:%.2f%%)" % (accuracy_total*100, accuracy_tooneg*100, accuracy_toopos*100))
171 | else:
172 | accuracy = numpy.mean((self.Y - pr_Y)**2)
173 | print("MSE: %.2f" % (numpy.sqrt(accuracy)))
174 |
175 | #save predicted result
176 | if self.ofp != False:
177 | numpy.savetxt(self.ofp, pr_Y, fmt='%5s')
178 |
179 | #----------------------------
180 | # select mode
181 | #----------------------------
182 | m = MakeModel(args)
183 | if args.mode == ['create']:
184 | #make model
185 | m.load_dataset()
186 | m.train_model()
187 | else:
188 | #predict dataset
189 | m.load_dataset()
190 | m.predict_ds()
191 |
192 |
193 |
--------------------------------------------------------------------------------
/is_future_purchase.ipynb:
--------------------------------------------------------------------------------
1 | {
2 | "cells": [
3 | {
4 | "cell_type": "code",
5 | "execution_count": 92,
6 | "metadata": {},
7 | "outputs": [
8 | {
9 | "data": {
10 | "text/html": [
11 | "\n",
12 | "
\n",
13 | " \n",
14 | " \n",
15 | " | \n",
16 | " Y | \n",
17 | " CustomerID | \n",
18 | " P1 | \n",
19 | " P2 | \n",
20 | " P3 | \n",
21 | " P4 | \n",
22 | " P5 | \n",
23 | " P6 | \n",
24 | " P7 | \n",
25 | " P8 | \n",
26 | " ... | \n",
27 | " P91 | \n",
28 | " P92 | \n",
29 | " P93 | \n",
30 | " P94 | \n",
31 | " P95 | \n",
32 | " P96 | \n",
33 | " P97 | \n",
34 | " P98 | \n",
35 | " P99 | \n",
36 | " P100 | \n",
37 | "
\n",
38 | " \n",
39 | " \n",
40 | " \n",
41 | " 0 | \n",
42 | " 0 | \n",
43 | " 12346 | \n",
44 | " 604 | \n",
45 | " 604.0 | \n",
46 | " NaN | \n",
47 | " NaN | \n",
48 | " NaN | \n",
49 | " NaN | \n",
50 | " NaN | \n",
51 | " NaN | \n",
52 | " ... | \n",
53 | " NaN | \n",
54 | " NaN | \n",
55 | " NaN | \n",
56 | " NaN | \n",
57 | " NaN | \n",
58 | " NaN | \n",
59 | " NaN | \n",
60 | " NaN | \n",
61 | " NaN | \n",
62 | " NaN | \n",
63 | "
\n",
64 | " \n",
65 | " 1 | \n",
66 | " 0 | \n",
67 | " 12348 | \n",
68 | " 120 | \n",
69 | " 238.0 | \n",
70 | " 60.0 | \n",
71 | " 60.0 | \n",
72 | " 109.0 | \n",
73 | " 109.0 | \n",
74 | " 181.0 | \n",
75 | " 553.0 | \n",
76 | " ... | \n",
77 | " NaN | \n",
78 | " NaN | \n",
79 | " NaN | \n",
80 | " NaN | \n",
81 | " NaN | \n",
82 | " NaN | \n",
83 | " NaN | \n",
84 | " NaN | \n",
85 | " NaN | \n",
86 | " NaN | \n",
87 | "
\n",
88 | " \n",
89 | " 2 | \n",
90 | " 0 | \n",
91 | " 12349 | \n",
92 | " 603 | \n",
93 | " 2259.0 | \n",
94 | " 1333.0 | \n",
95 | " 649.0 | \n",
96 | " 1072.0 | \n",
97 | " 488.0 | \n",
98 | " 344.0 | \n",
99 | " 140.0 | \n",
100 | " ... | \n",
101 | " NaN | \n",
102 | " NaN | \n",
103 | " NaN | \n",
104 | " NaN | \n",
105 | " NaN | \n",
106 | " NaN | \n",
107 | " NaN | \n",
108 | " NaN | \n",
109 | " NaN | \n",
110 | " NaN | \n",
111 | "
\n",
112 | " \n",
113 | " 3 | \n",
114 | " 0 | \n",
115 | " 12350 | \n",
116 | " 217 | \n",
117 | " 1117.0 | \n",
118 | " 1073.0 | \n",
119 | " 1847.0 | \n",
120 | " 384.0 | \n",
121 | " 2139.0 | \n",
122 | " 102.0 | \n",
123 | " 148.0 | \n",
124 | " ... | \n",
125 | " NaN | \n",
126 | " NaN | \n",
127 | " NaN | \n",
128 | " NaN | \n",
129 | " NaN | \n",
130 | " NaN | \n",
131 | " NaN | \n",
132 | " NaN | \n",
133 | " NaN | \n",
134 | " NaN | \n",
135 | "
\n",
136 | " \n",
137 | " 4 | \n",
138 | " 0 | \n",
139 | " 12352 | \n",
140 | " 1155 | \n",
141 | " 630.0 | \n",
142 | " 779.0 | \n",
143 | " 1325.0 | \n",
144 | " 402.0 | \n",
145 | " 348.0 | \n",
146 | " 524.0 | \n",
147 | " 1839.0 | \n",
148 | " ... | \n",
149 | " 749.0 | \n",
150 | " 537.0 | \n",
151 | " 1066.0 | \n",
152 | " 1196.0 | \n",
153 | " 11.0 | \n",
154 | " NaN | \n",
155 | " NaN | \n",
156 | " NaN | \n",
157 | " NaN | \n",
158 | " NaN | \n",
159 | "
\n",
160 | " \n",
161 | "
\n",
162 | "
5 rows × 102 columns
\n",
163 | "
"
164 | ],
165 | "text/plain": [
166 | " Y CustomerID P1 P2 P3 P4 P5 P6 P7 P8 \\\n",
167 | "0 0 12346 604 604.0 NaN NaN NaN NaN NaN NaN \n",
168 | "1 0 12348 120 238.0 60.0 60.0 109.0 109.0 181.0 553.0 \n",
169 | "2 0 12349 603 2259.0 1333.0 649.0 1072.0 488.0 344.0 140.0 \n",
170 | "3 0 12350 217 1117.0 1073.0 1847.0 384.0 2139.0 102.0 148.0 \n",
171 | "4 0 12352 1155 630.0 779.0 1325.0 402.0 348.0 524.0 1839.0 \n",
172 | "\n",
173 | " ... P91 P92 P93 P94 P95 P96 P97 P98 P99 P100 \n",
174 | "0 ... NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN \n",
175 | "1 ... NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN \n",
176 | "2 ... NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN \n",
177 | "3 ... NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN \n",
178 | "4 ... 749.0 537.0 1066.0 1196.0 11.0 NaN NaN NaN NaN NaN \n",
179 | "\n",
180 | "[5 rows x 102 columns]"
181 | ]
182 | },
183 | "execution_count": 92,
184 | "metadata": {},
185 | "output_type": "execute_result"
186 | }
187 | ],
188 | "source": [
189 | "import numpy\n",
190 | "from keras.models import Sequential\n",
191 | "from keras.layers import Dense\n",
192 | "from keras.layers import LSTM\n",
193 | "from keras.layers.embeddings import Embedding\n",
194 | "from keras.preprocessing import sequence\n",
195 | "import pandas as pd\n",
196 | "from sklearn import cross_validation\n",
197 | "\n",
198 | "# load data\n",
199 | "top_words = 4214 # insert max index of items + 1\n",
200 | "max_length = 10 # length of sequential data\n",
201 | "df = pd.read_csv(\"c:/dev/dl/purchacedata_base.csv\", header=0)\n",
202 | "\n",
203 | "df.head()"
204 | ]
205 | },
206 | {
207 | "cell_type": "code",
208 | "execution_count": 93,
209 | "metadata": {},
210 | "outputs": [
211 | {
212 | "name": "stdout",
213 | "output_type": "stream",
214 | "text": [
215 | "(2835, 10)\n",
216 | "(946, 10)\n",
217 | "[0 0 0 0 0 1 1 1 1 0 0 0 0 1 0 0 1 1 0 1 0 0 0 0 1 0 0 0 1 0 0 1 0 0 1 0 0\n",
218 | " 0 0 0 0 0 1 1 1 0 0 1 0 1 1 1 1 1 1 1 0 1 1 0 1 0 0 0 1 1 1 0 1 0 0 1 0 0\n",
219 | " 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 1 0 1 0 0 0 1 0 0 0 1 0 0 0 0 0 0 0 1 0 0\n",
220 | " 1 1 1 0 0 1 0 0 0 0 0 0 0 0 0 1 0 0 0 0 1 0 0 1 0 0 0 0 0 1 0 0 0 1 1 0 0\n",
221 | " 1 0 0 0 0 1 1 0 0 0 0 0 0 0 0 0 0 1 0 1 1 0 0 1 0 0 0 0 0 1 0 0 0 1 0 0 0\n",
222 | " 0 1 0 0 0 0 1 0 0 1 0 0 0 0 0 1 0 0 0 0 1 0 1 0 0 0 1 0 0 0 1 0 1 0 0 0 0\n",
223 | " 1 0 1 0 1 0 0 0 0 1 0 0 0 0 1 0 1 1 1 0 0 0 0 0 0 0 1 0 1 0 0 0 0 0 1 0 1\n",
224 | " 1 0 1 1 0 0 0 0 0 0 0 0 0 1 0 0 0 0 1 0 0 1 0 0 0 0 0 1 1 0 1 1 0 1 1 0 0\n",
225 | " 0 1 0 0 1 0 1 0 0 0 0 0 0 0 0 0 1 1 0 0 0 1 0 0 1 0 0 0 1 1 1 1 0 0 1 1 1\n",
226 | " 0 0 0 0 0 0 1 0 0 0 0 0 1 0 1 0 1 0 1 0 0 0 0 1 0 0 0 1 0 0 1 0 1 0 1 0 1\n",
227 | " 1 0 1 0 0 0 1 0 0 0 1 0 0 0 1 1 0 1 0 1 1 1 1 0 0 0 0 1 0 0 0 0 1 0 0 0 1\n",
228 | " 0 1 1 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 1 1 0 1 0 1 0 1 0\n",
229 | " 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 1 1 0 0 1 0 0 1 0 0\n",
230 | " 0 0 0 0 0 0 0 0 0 1 0 0 1 0 0 0 0 0 0 1 0 1 0 1 0 0 1 0 0 1 0 1 0 0 0 0 1\n",
231 | " 0 1 0 0 0 1 1 1 0 0 0 0 0 0 1 1 1 0 1 1 0 0 0 0 0 0 1 0 0 1 0 0 0 1 0 0 1\n",
232 | " 0 0 0 0 1 0 0 0 0 0 1 0 0 0 0 0 0 0 1 0 0 0 1 0 0 0 0 0 0 1 0 1 0 1 0 0 0\n",
233 | " 0 1 1 0 0 0 0 0 0 1 1 0 1 1 0 0 1 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 1 0 0 1 0\n",
234 | " 0 0 0 1 0 0 0 0 0 0 0 0 1 0 0 0 0 1 1 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0\n",
235 | " 0 0 0 0 0 0 0 0 0 0 0 1 1 0 1 0 1 1 1 0 1 1 0 1 0 1 0 0 0 1 0 0 0 0 0 0 0\n",
236 | " 0 0 0 0 1 0 0 0 1 0 0 0 0 1 0 0 0 1 0 1 0 0 1 1 0 0 0 0 0 0 0 0 0 0 0 1 0\n",
237 | " 0 1 0 0 0 1 0 1 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 1 0 0 0 0 1 1 0 0 0 0 0 0\n",
238 | " 0 0 0 0 0 0 0 0 1 0 0 1 0 1 0 0 0 0 0 1 0 0 0 0 0 0 0 1 0 1 0 0 0 1 1 0 0\n",
239 | " 1 1 0 0 0 0 0 1 0 1 0 1 0 0 0 0 0 0 0 0 1 1 0 0 0 1 0 1 0 0 0 0 1 0 0 1 1\n",
240 | " 0 1 0 1 1 0 0 1 1 1 1 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 1 0 1 0 0 0 0 0 1 0\n",
241 | " 0 0 0 1 0 0 1 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 1 1 0 0 0 0 0 1 1 0 0 0 0 1\n",
242 | " 1 1 0 1 0 1 1 0 0 0 0 1 0 0 0 0 1 0 1 0 1]\n"
243 | ]
244 | }
245 | ],
246 | "source": [
247 | "# extract columns and drop rows having NaN\n",
248 | "df = df[df.columns[0:(max_length + 2)]].dropna().astype(int)\n",
249 | "\n",
250 | "train_X, test_X, train_Y, test_Y = cross_validation.train_test_split(df[df.columns[2:(max_length + 2)]], df[\"Y\"])\n",
251 | "\n",
252 | "train_X = train_X.values\n",
253 | "train_Y = train_Y.values\n",
254 | "test_X = test_X.values\n",
255 | "test_Y = test_Y.values\n",
256 | "\n",
257 | "print(train_X.shape)\n",
258 | "print(test_X.shape)\n",
259 | "print(test_Y)"
260 | ]
261 | },
262 | {
263 | "cell_type": "code",
264 | "execution_count": 107,
265 | "metadata": {},
266 | "outputs": [
267 | {
268 | "name": "stdout",
269 | "output_type": "stream",
270 | "text": [
271 | "_________________________________________________________________\n",
272 | "Layer (type) Output Shape Param # \n",
273 | "=================================================================\n",
274 | "embedding_29 (Embedding) (None, 10, 5) 21070 \n",
275 | "_________________________________________________________________\n",
276 | "lstm_29 (LSTM) (None, 3) 108 \n",
277 | "_________________________________________________________________\n",
278 | "dense_29 (Dense) (None, 1) 4 \n",
279 | "=================================================================\n",
280 | "Total params: 21,182\n",
281 | "Trainable params: 21,182\n",
282 | "Non-trainable params: 0\n",
283 | "_________________________________________________________________\n",
284 | "None\n",
285 | "Train on 2835 samples, validate on 946 samples\n",
286 | "Epoch 1/3\n",
287 | "2835/2835 [==============================] - 1s - loss: 0.6792 - acc: 0.6942 - val_loss: 0.6594 - val_acc: 0.7241\n",
288 | "Epoch 2/3\n",
289 | "2835/2835 [==============================] - 0s - loss: 0.6411 - acc: 0.7012 - val_loss: 0.6106 - val_acc: 0.7241\n",
290 | "Epoch 3/3\n",
291 | "2835/2835 [==============================] - 0s - loss: 0.6002 - acc: 0.7012 - val_loss: 0.5859 - val_acc: 0.7241\n"
292 | ]
293 | },
294 | {
295 | "data": {
296 | "text/plain": [
297 | ""
298 | ]
299 | },
300 | "execution_count": 107,
301 | "metadata": {},
302 | "output_type": "execute_result"
303 | }
304 | ],
305 | "source": [
306 | "# create the model\n",
307 | "embedding_vector_length = 5\n",
308 | "model = Sequential()\n",
309 | "model.add(Embedding(top_words, embedding_vector_length, input_length=max_length))\n",
310 | "model.add(LSTM(3))\n",
311 | "model.add(Dense(1, activation='sigmoid'))\n",
312 | "model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])\n",
313 | "print(model.summary())\n",
314 | "model.fit(train_X, train_Y, validation_data=(test_X, test_Y), epochs=3, batch_size=64)"
315 | ]
316 | },
317 | {
318 | "cell_type": "code",
319 | "execution_count": 108,
320 | "metadata": {},
321 | "outputs": [
322 | {
323 | "name": "stdout",
324 | "output_type": "stream",
325 | "text": [
326 | "Accuracy: 72.41%\n"
327 | ]
328 | }
329 | ],
330 | "source": [
331 | "# Final evaluation of the model\n",
332 | "scores = model.evaluate(test_X, test_Y, verbose=0)\n",
333 | "print(\"Accuracy: %.2f%%\" % (scores[1]*100))"
334 | ]
335 | },
336 | {
337 | "cell_type": "code",
338 | "execution_count": null,
339 | "metadata": {
340 | "collapsed": true
341 | },
342 | "outputs": [],
343 | "source": []
344 | }
345 | ],
346 | "metadata": {
347 | "kernelspec": {
348 | "display_name": "Python 3",
349 | "language": "python",
350 | "name": "python3"
351 | },
352 | "language_info": {
353 | "codemirror_mode": {
354 | "name": "ipython",
355 | "version": 3
356 | },
357 | "file_extension": ".py",
358 | "mimetype": "text/x-python",
359 | "name": "python",
360 | "nbconvert_exporter": "python",
361 | "pygments_lexer": "ipython3",
362 | "version": "3.5.0"
363 | }
364 | },
365 | "nbformat": 4,
366 | "nbformat_minor": 2
367 | }
368 |
--------------------------------------------------------------------------------
/tokyo-weather-2003-2012.csv:
--------------------------------------------------------------------------------
1 | ice_sales,year,month,avg_temp,total_rain,humidity,num_day_over25deg
2 | 331,2003,1,9.3,101,46,0
3 | 268,2003,2,9.9,53.5,52,0
4 | 365,2003,3,12.7,159.5,49,0
5 | 492,2003,4,19.2,121,61,3
6 | 632,2003,5,22.4,172.5,65,7
7 | 730,2003,6,26.6,85,69,21
8 | 821,2003,7,26,187.5,75,21
9 | 1057,2003,8,29.5,370,73,26
10 | 724,2003,9,28.1,150,66,23
11 | 430,2003,10,21.4,171.5,59,3
12 | 363,2003,11,17.4,229.5,67,0
13 | 415,2003,12,13.2,53,50,0
14 | 351,2004,1,10.1,3.5,43,0
15 | 322,2004,2,12.9,20,45,0
16 | 367,2004,3,14,129.5,53,0
17 | 508,2004,4,21.3,69.5,51,3
18 | 667,2004,5,23.7,149,67,13
19 | 772,2004,6,27.5,112.5,66,24
20 | 1148,2004,7,33.1,23.5,62,31
21 | 1080,2004,8,31,79.5,65,28
22 | 653,2004,9,28.7,195,68,26
23 | 434,2004,10,20.7,780,69,3
24 | 358,2004,11,19,108.5,60,0
25 | 388,2004,12,13.4,79.5,49,0
26 | 323,2005,1,10,77,47,0
27 | 283,2005,2,9.9,48,45,0
28 | 357,2005,3,13.1,71,49,0
29 | 543,2005,4,19.6,81,54,2
30 | 667,2005,5,21.9,180.5,58,6
31 | 812,2005,6,26.7,170.5,70,22
32 | 1037,2005,7,29.1,247.5,71,27
33 | 1179,2005,8,31.8,189.5,68,31
34 | 739,2005,9,28.2,177.5,67,24
35 | 453,2005,10,22.3,201.5,69,7
36 | 315,2005,11,17,34.5,52,0
37 | 359,2005,12,10.2,3.5,39,0
38 | 322,2006,1,8.5,67,44,0
39 | 279,2006,2,10.5,113,53,0
40 | 373,2006,3,14,79.5,48,0
41 | 457,2006,4,17.5,123,57,0
42 | 602,2006,5,22.7,99,65,7
43 | 748,2006,6,25.4,138.5,71,15
44 | 973,2006,7,28.6,165,74,26
45 | 1193,2006,8,31.1,126,69,31
46 | 654,2006,9,27.1,175.5,68,20
47 | 493,2006,10,22.9,318,66,9
48 | 336,2006,11,17.9,135,59,0
49 | 392,2006,12,12.5,200.5,52,0
50 | 347,2007,1,10.9,42,45,0
51 | 292,2007,2,12.8,57,45,0
52 | 387,2007,3,15,77,44,0
53 | 466,2007,4,17.9,134,59,1
54 | 652,2007,5,24,115.5,58,13
55 | 768,2007,6,27.1,80,66,23
56 | 908,2007,7,27.4,253,74,26
57 | 1279,2007,8,33,9.5,66,31
58 | 784,2007,9,28.5,319.5,71,25
59 | 469,2007,10,22.2,135.5,63,4
60 | 324,2007,11,16.6,37,56,0
61 | 405,2007,12,12.6,72,54,0
62 | 346,2008,1,9.4,17.5,45,0
63 | 288,2008,2,9.6,57,41,0
64 | 404,2008,3,14.9,119.5,52,0
65 | 501,2008,4,18.4,240,59,1
66 | 689,2008,5,22,255,65,8
67 | 727,2008,6,24.6,225.5,72,15
68 | 1182,2008,7,30.9,48,71,31
69 | 1190,2008,8,30.7,387.5,74,28
70 | 691,2008,9,27.7,158.5,69,25
71 | 477,2008,10,22.6,204.5,66,1
72 | 355,2008,11,16.4,74,56,0
73 | 414,2008,12,13.7,70.5,53,0
74 | 351,2009,1,10.2,142,48,0
75 | 303,2009,2,11.5,46.5,50,0
76 | 386,2009,3,13.7,98.5,48,0
77 | 569,2009,4,20.2,162.5,54,1
78 | 768,2009,5,23.6,242,64,9
79 | 819,2009,6,25.8,226,72,16
80 | 1072,2009,7,29.3,78.5,72,28
81 | 1215,2009,8,30.1,242,69,30
82 | 723,2009,9,26.5,53,64,21
83 | 495,2009,10,22.3,276.5,64,5
84 | 402,2009,11,17,151.5,63,1
85 | 440,2009,12,12.4,82.5,51,0
86 | 362,2010,1,11,9,41,0
87 | 305,2010,2,9.9,115,60,0
88 | 383,2010,3,13.2,143.5,61,0
89 | 464,2010,4,16.6,214,62,1
90 | 752,2010,5,23,114,60,8
91 | 841,2010,6,27.5,108,67,24
92 | 1211,2010,7,31.6,70,70,31
93 | 1451,2010,8,33.5,27,67,31
94 | 864,2010,9,29,428,68,22
95 | 504,2010,10,21.8,211,68,6
96 | 351,2010,11,17.2,94.5,56,0
97 | 423,2010,12,13.7,145.5,50,0
98 | 346,2011,1,9.1,3.5,36,0
99 | 289,2011,2,11.2,151,52,0
100 | 329,2011,3,12.3,74,47,0
101 | 462,2011,4,18.9,96,50,0
102 | 672,2011,5,22.2,213.5,63,8
103 | 791,2011,6,26,116.5,71,15
104 | 1265,2011,7,30.9,54.5,67,29
105 | 1241,2011,8,31.2,244,71,29
106 | 767,2011,9,28.8,235,68,23
107 | 516,2011,10,23,119.5,61,9
108 | 393,2011,11,18.3,112.5,58,0
109 | 423,2011,12,11.1,59.5,48,0
110 | 339,2012,1,8.3,50,43,0
111 | 274,2012,2,9.1,94,49,0
112 | 385,2012,3,12.5,144.5,59,0
113 | 524,2012,4,18.5,118.5,63,1
114 | 671,2012,5,23.6,231,65,12
115 | 798,2012,6,24.8,185,73,15
116 | 1165,2012,7,30.1,130,75,27
117 | 1332,2012,8,33.1,25,69,31
118 | 849,2012,9,29.8,214.5,73,27
119 | 515,2012,10,23,154.5,65,9
120 | 326,2012,11,16.3,154,58,0
121 | 414,2012,12,11.2,69,52,0
122 |
--------------------------------------------------------------------------------
/転移学習テスト.ipynb:
--------------------------------------------------------------------------------
1 | {
2 | "cells": [
3 | {
4 | "cell_type": "code",
5 | "execution_count": 1,
6 | "metadata": {},
7 | "outputs": [
8 | {
9 | "name": "stderr",
10 | "output_type": "stream",
11 | "text": [
12 | "Using TensorFlow backend.\n"
13 | ]
14 | },
15 | {
16 | "name": "stdout",
17 | "output_type": "stream",
18 | "text": [
19 | "(120, 4) (30, 4) (120, 3) (30, 3)\n"
20 | ]
21 | }
22 | ],
23 | "source": [
24 | "import numpy as np\n",
25 | "\n",
26 | "from sklearn import datasets\n",
27 | "from sklearn.model_selection import train_test_split\n",
28 | "\n",
29 | "from keras.models import Sequential\n",
30 | "from keras.layers.core import Dense, Activation\n",
31 | "from keras.layers import Dense, Dropout, Flatten\n",
32 | "from keras.utils import np_utils\n",
33 | "from sklearn import preprocessing\n",
34 | "import matplotlib.pyplot as plt\n",
35 | "\n",
36 | "a = l = []\n",
37 | "\n",
38 | "def build_multilayer_perceptron():\n",
39 | " \"\"\"多層パーセプトロンモデルを構築\"\"\"\n",
40 | " model = Sequential()\n",
41 | " model.add(Dense(4, input_shape=(4, ), name='l1'))\n",
42 | " model.add(Activation('relu'))\n",
43 | " model.add(Dense(4, input_shape=(4, ), name='l2'))\n",
44 | " model.add(Activation('relu'))\n",
45 | " #model.add(Dropout(0.5))\n",
46 | " model.add(Dense(3, name='cls'))\n",
47 | " model.add(Activation('softmax'))\n",
48 | " return model\n",
49 | "\n",
50 | "# Irisデータをロード\n",
51 | "iris = datasets.load_iris()\n",
52 | "X = iris.data\n",
53 | "Y = iris.target\n",
54 | "\n",
55 | "# データの標準化\n",
56 | "X = preprocessing.scale(X)\n",
57 | "\n",
58 | "# ラベルをone-hot-encoding形式に変換\n",
59 | "Y = np_utils.to_categorical(Y)\n",
60 | "\n",
61 | "# 訓練データとテストデータに分割\n",
62 | "train_X, test_X, train_Y, test_Y = train_test_split(X, Y, train_size=0.8)\n",
63 | "print(train_X.shape, test_X.shape, train_Y.shape, test_Y.shape)"
64 | ]
65 | },
66 | {
67 | "cell_type": "code",
68 | "execution_count": 2,
69 | "metadata": {},
70 | "outputs": [
71 | {
72 | "name": "stdout",
73 | "output_type": "stream",
74 | "text": [
75 | "Epoch 1/50\n",
76 | "120/120 [==============================] - 0s - loss: 1.2139 - acc: 0.3833 \n",
77 | "Epoch 2/50\n",
78 | "120/120 [==============================] - 0s - loss: 1.1321 - acc: 0.5250 \n",
79 | "Epoch 3/50\n",
80 | "120/120 [==============================] - 0s - loss: 1.0962 - acc: 0.5917 \n",
81 | "Epoch 4/50\n",
82 | "120/120 [==============================] - 0s - loss: 1.0738 - acc: 0.6083 \n",
83 | "Epoch 5/50\n",
84 | "120/120 [==============================] - 0s - loss: 1.0572 - acc: 0.6000 \n",
85 | "Epoch 6/50\n",
86 | "120/120 [==============================] - 0s - loss: 1.0434 - acc: 0.6250 \n",
87 | "Epoch 7/50\n",
88 | "120/120 [==============================] - 0s - loss: 1.0319 - acc: 0.6333 \n",
89 | "Epoch 8/50\n",
90 | "120/120 [==============================] - 0s - loss: 1.0205 - acc: 0.6333 \n",
91 | "Epoch 9/50\n",
92 | "120/120 [==============================] - 0s - loss: 1.0094 - acc: 0.6333 \n",
93 | "Epoch 10/50\n",
94 | "120/120 [==============================] - 0s - loss: 0.9990 - acc: 0.6250 \n",
95 | "Epoch 11/50\n",
96 | "120/120 [==============================] - 0s - loss: 0.9877 - acc: 0.6250 \n",
97 | "Epoch 12/50\n",
98 | "120/120 [==============================] - 0s - loss: 0.9764 - acc: 0.6333 \n",
99 | "Epoch 13/50\n",
100 | "120/120 [==============================] - 0s - loss: 0.9639 - acc: 0.6250 \n",
101 | "Epoch 14/50\n",
102 | "120/120 [==============================] - 0s - loss: 0.9502 - acc: 0.6333 \n",
103 | "Epoch 15/50\n",
104 | "120/120 [==============================] - 0s - loss: 0.9354 - acc: 0.6417 \n",
105 | "Epoch 16/50\n",
106 | "120/120 [==============================] - 0s - loss: 0.9204 - acc: 0.6583 \n",
107 | "Epoch 17/50\n",
108 | "120/120 [==============================] - 0s - loss: 0.9053 - acc: 0.6417 \n",
109 | "Epoch 18/50\n",
110 | "120/120 [==============================] - 0s - loss: 0.8875 - acc: 0.6500 \n",
111 | "Epoch 19/50\n",
112 | "120/120 [==============================] - 0s - loss: 0.8714 - acc: 0.6500 \n",
113 | "Epoch 20/50\n",
114 | "120/120 [==============================] - 0s - loss: 0.8537 - acc: 0.6500 \n",
115 | "Epoch 21/50\n",
116 | "120/120 [==============================] - 0s - loss: 0.8368 - acc: 0.6417 \n",
117 | "Epoch 22/50\n",
118 | "120/120 [==============================] - 0s - loss: 0.8197 - acc: 0.6417 \n",
119 | "Epoch 23/50\n",
120 | "120/120 [==============================] - 0s - loss: 0.8035 - acc: 0.6333 \n",
121 | "Epoch 24/50\n",
122 | "120/120 [==============================] - 0s - loss: 0.7868 - acc: 0.6250 \n",
123 | "Epoch 25/50\n",
124 | "120/120 [==============================] - 0s - loss: 0.7716 - acc: 0.6250 \n",
125 | "Epoch 26/50\n",
126 | "120/120 [==============================] - 0s - loss: 0.7556 - acc: 0.6167 \n",
127 | "Epoch 27/50\n",
128 | "120/120 [==============================] - 0s - loss: 0.7407 - acc: 0.6167 \n",
129 | "Epoch 28/50\n",
130 | "120/120 [==============================] - 0s - loss: 0.7257 - acc: 0.6167 \n",
131 | "Epoch 29/50\n",
132 | "120/120 [==============================] - 0s - loss: 0.7129 - acc: 0.5917 \n",
133 | "Epoch 30/50\n",
134 | "120/120 [==============================] - 0s - loss: 0.7001 - acc: 0.5583 \n",
135 | "Epoch 31/50\n",
136 | "120/120 [==============================] - 0s - loss: 0.6875 - acc: 0.5667 \n",
137 | "Epoch 32/50\n",
138 | "120/120 [==============================] - 0s - loss: 0.6759 - acc: 0.5333 \n",
139 | "Epoch 33/50\n",
140 | "120/120 [==============================] - 0s - loss: 0.6651 - acc: 0.5417 \n",
141 | "Epoch 34/50\n",
142 | "120/120 [==============================] - 0s - loss: 0.6552 - acc: 0.5500 \n",
143 | "Epoch 35/50\n",
144 | "120/120 [==============================] - 0s - loss: 0.6440 - acc: 0.5167 \n",
145 | "Epoch 36/50\n",
146 | "120/120 [==============================] - 0s - loss: 0.6345 - acc: 0.5167 \n",
147 | "Epoch 37/50\n",
148 | "120/120 [==============================] - 0s - loss: 0.6260 - acc: 0.5417 \n",
149 | "Epoch 38/50\n",
150 | "120/120 [==============================] - 0s - loss: 0.6176 - acc: 0.5500 \n",
151 | "Epoch 39/50\n",
152 | "120/120 [==============================] - 0s - loss: 0.6088 - acc: 0.5833 \n",
153 | "Epoch 40/50\n",
154 | "120/120 [==============================] - 0s - loss: 0.6002 - acc: 0.6000 \n",
155 | "Epoch 41/50\n",
156 | "120/120 [==============================] - 0s - loss: 0.5923 - acc: 0.6417 \n",
157 | "Epoch 42/50\n",
158 | "120/120 [==============================] - 0s - loss: 0.5868 - acc: 0.6417 \n",
159 | "Epoch 43/50\n",
160 | "120/120 [==============================] - 0s - loss: 0.5771 - acc: 0.6500 \n",
161 | "Epoch 44/50\n",
162 | "120/120 [==============================] - 0s - loss: 0.5702 - acc: 0.6500 \n",
163 | "Epoch 45/50\n",
164 | "120/120 [==============================] - 0s - loss: 0.5627 - acc: 0.6500 \n",
165 | "Epoch 46/50\n",
166 | "120/120 [==============================] - 0s - loss: 0.5563 - acc: 0.6500 \n",
167 | "Epoch 47/50\n",
168 | "120/120 [==============================] - 0s - loss: 0.5495 - acc: 0.6500 \n",
169 | "Epoch 48/50\n",
170 | "120/120 [==============================] - 0s - loss: 0.5372 - acc: 0.6500 \n",
171 | "Epoch 49/50\n",
172 | "120/120 [==============================] - 0s - loss: 0.5218 - acc: 0.6500 \n",
173 | "Epoch 50/50\n",
174 | "120/120 [==============================] - 0s - loss: 0.4948 - acc: 0.6500 \n",
175 | "Accuracy = 0.73\n"
176 | ]
177 | }
178 | ],
179 | "source": [
180 | "# モデル構築\n",
181 | "model = build_multilayer_perceptron()\n",
182 | "model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['accuracy'])\n",
183 | "\n",
184 | "# モデル訓練\n",
185 | "model.fit(train_X, train_Y, epochs=50, batch_size=1, verbose=1)\n",
186 | "\n",
187 | "# モデル評価\n",
188 | "loss, accuracy = model.evaluate(test_X, test_Y, verbose=0)\n",
189 | "print(\"Accuracy = {:.2f}\".format(accuracy))\n"
190 | ]
191 | },
192 | {
193 | "cell_type": "code",
194 | "execution_count": 21,
195 | "metadata": {
196 | "collapsed": true
197 | },
198 | "outputs": [],
199 | "source": [
200 | "#モデル保存\n",
201 | "model.save_weights('my_model_weights.h5')\n"
202 | ]
203 | },
204 | {
205 | "cell_type": "code",
206 | "execution_count": 4,
207 | "metadata": {},
208 | "outputs": [
209 | {
210 | "name": "stdout",
211 | "output_type": "stream",
212 | "text": [
213 | "_________________________________________________________________\n",
214 | "Layer (type) Output Shape Param # \n",
215 | "=================================================================\n",
216 | "l1 (Dense) (None, 4) 20 \n",
217 | "_________________________________________________________________\n",
218 | "activation_7 (Activation) (None, 4) 0 \n",
219 | "_________________________________________________________________\n",
220 | "l2 (Dense) (None, 4) 20 \n",
221 | "_________________________________________________________________\n",
222 | "activation_8 (Activation) (None, 4) 0 \n",
223 | "_________________________________________________________________\n",
224 | "dropout_2 (Dropout) (None, 4) 0 \n",
225 | "_________________________________________________________________\n",
226 | "cls (Dense) (None, 3) 15 \n",
227 | "_________________________________________________________________\n",
228 | "activation_9 (Activation) (None, 3) 0 \n",
229 | "=================================================================\n",
230 | "Total params: 55\n",
231 | "Trainable params: 55\n",
232 | "Non-trainable params: 0\n",
233 | "_________________________________________________________________\n"
234 | ]
235 | }
236 | ],
237 | "source": [
238 | "#モデルのロード\n",
239 | "model = Sequential()\n",
240 | "model.add(Dense(4, input_shape=(4, ), name='l1'))\n",
241 | "model.add(Activation('relu'))\n",
242 | "model.add(Dense(4, input_shape=(4, ), name='l2'))\n",
243 | "model.add(Activation('relu'))\n",
244 | "model.add(Dropout(0.5))\n",
245 | "model.add(Dense(3, name='cls'))\n",
246 | "model.add(Activation('softmax'))\n",
247 | "\n",
248 | "model.load_weights('my_model_weights.h5', by_name=True)\n",
249 | "model.summary()\n",
250 | "\n",
251 | "#\n",
252 | "#for layer in model.layers[:2]:\n",
253 | "# layer.trainable = False"
254 | ]
255 | },
256 | {
257 | "cell_type": "code",
258 | "execution_count": 5,
259 | "metadata": {},
260 | "outputs": [
261 | {
262 | "name": "stdout",
263 | "output_type": "stream",
264 | "text": [
265 | "Epoch 1/50\n",
266 | "30/30 [==============================] - 0s - loss: 1.9885 - acc: 0.5333 \n",
267 | "Epoch 2/50\n",
268 | "30/30 [==============================] - 0s - loss: 1.2767 - acc: 0.6667 \n",
269 | "Epoch 3/50\n",
270 | "30/30 [==============================] - 0s - loss: 0.4599 - acc: 0.7667 \n",
271 | "Epoch 4/50\n",
272 | "30/30 [==============================] - 0s - loss: 1.4425 - acc: 0.6333 \n",
273 | "Epoch 5/50\n",
274 | "30/30 [==============================] - 0s - loss: 1.2697 - acc: 0.7000 \n",
275 | "Epoch 6/50\n",
276 | "30/30 [==============================] - 0s - loss: 0.9079 - acc: 0.7667 \n",
277 | "Epoch 7/50\n",
278 | "30/30 [==============================] - 0s - loss: 1.2881 - acc: 0.6667 \n",
279 | "Epoch 8/50\n",
280 | "30/30 [==============================] - 0s - loss: 1.5422 - acc: 0.6333 \n",
281 | "Epoch 9/50\n",
282 | "30/30 [==============================] - 0s - loss: 0.5990 - acc: 0.8333 \n",
283 | "Epoch 10/50\n",
284 | "30/30 [==============================] - 0s - loss: 1.8923 - acc: 0.5333 \n",
285 | "Epoch 11/50\n",
286 | "30/30 [==============================] - 0s - loss: 1.1823 - acc: 0.7333 \n",
287 | "Epoch 12/50\n",
288 | "30/30 [==============================] - 0s - loss: 1.2670 - acc: 0.6333 \n",
289 | "Epoch 13/50\n",
290 | "30/30 [==============================] - 0s - loss: 0.8398 - acc: 0.7333 \n",
291 | "Epoch 14/50\n",
292 | "30/30 [==============================] - 0s - loss: 0.9550 - acc: 0.7333 \n",
293 | "Epoch 15/50\n",
294 | "30/30 [==============================] - 0s - loss: 0.9578 - acc: 0.7333 \n",
295 | "Epoch 16/50\n",
296 | "30/30 [==============================] - ETA: 0s - loss: 0.0011 - acc: 1.000 - 0s - loss: 1.0942 - acc: 0.7000 \n",
297 | "Epoch 17/50\n",
298 | "30/30 [==============================] - 0s - loss: 0.6804 - acc: 0.7333 \n",
299 | "Epoch 18/50\n",
300 | "30/30 [==============================] - 0s - loss: 0.8193 - acc: 0.7667 \n",
301 | "Epoch 19/50\n",
302 | "30/30 [==============================] - 0s - loss: 1.0397 - acc: 0.7333 \n",
303 | "Epoch 20/50\n",
304 | "30/30 [==============================] - 0s - loss: 1.1131 - acc: 0.6333 \n",
305 | "Epoch 21/50\n",
306 | "30/30 [==============================] - 0s - loss: 0.7545 - acc: 0.7667 \n",
307 | "Epoch 22/50\n",
308 | "30/30 [==============================] - 0s - loss: 0.5940 - acc: 0.8333 \n",
309 | "Epoch 23/50\n",
310 | "30/30 [==============================] - 0s - loss: 0.7878 - acc: 0.7000 \n",
311 | "Epoch 24/50\n",
312 | "30/30 [==============================] - 0s - loss: 0.8876 - acc: 0.7667 \n",
313 | "Epoch 25/50\n",
314 | "30/30 [==============================] - 0s - loss: 0.8782 - acc: 0.7333 \n",
315 | "Epoch 26/50\n",
316 | "30/30 [==============================] - 0s - loss: 0.7679 - acc: 0.7333 \n",
317 | "Epoch 27/50\n",
318 | "30/30 [==============================] - 0s - loss: 0.8210 - acc: 0.7333 \n",
319 | "Epoch 28/50\n",
320 | "30/30 [==============================] - 0s - loss: 0.5197 - acc: 0.8333 \n",
321 | "Epoch 29/50\n",
322 | "30/30 [==============================] - 0s - loss: 0.7196 - acc: 0.7333 \n",
323 | "Epoch 30/50\n",
324 | "30/30 [==============================] - 0s - loss: 0.9312 - acc: 0.6667 \n",
325 | "Epoch 31/50\n",
326 | "30/30 [==============================] - 0s - loss: 0.4061 - acc: 0.8667 \n",
327 | "Epoch 32/50\n",
328 | "30/30 [==============================] - 0s - loss: 0.8994 - acc: 0.6667 \n",
329 | "Epoch 33/50\n",
330 | "30/30 [==============================] - 0s - loss: 0.5981 - acc: 0.8000 \n",
331 | "Epoch 34/50\n",
332 | "30/30 [==============================] - 0s - loss: 0.7241 - acc: 0.7333 \n",
333 | "Epoch 35/50\n",
334 | "30/30 [==============================] - 0s - loss: 0.8799 - acc: 0.7000 \n",
335 | "Epoch 36/50\n",
336 | "30/30 [==============================] - 0s - loss: 0.5681 - acc: 0.8000 \n",
337 | "Epoch 37/50\n",
338 | "30/30 [==============================] - 0s - loss: 0.8932 - acc: 0.6667 \n",
339 | "Epoch 38/50\n",
340 | "30/30 [==============================] - 0s - loss: 0.8078 - acc: 0.7333 \n",
341 | "Epoch 39/50\n",
342 | "30/30 [==============================] - 0s - loss: 0.9197 - acc: 0.7000 \n",
343 | "Epoch 40/50\n",
344 | "30/30 [==============================] - 0s - loss: 0.8454 - acc: 0.7667 \n",
345 | "Epoch 41/50\n",
346 | "30/30 [==============================] - 0s - loss: 0.6268 - acc: 0.8000 \n",
347 | "Epoch 42/50\n",
348 | "30/30 [==============================] - 0s - loss: 0.6760 - acc: 0.7333 \n",
349 | "Epoch 43/50\n",
350 | "30/30 [==============================] - 0s - loss: 1.0148 - acc: 0.6667 \n",
351 | "Epoch 44/50\n",
352 | "30/30 [==============================] - 0s - loss: 0.9666 - acc: 0.6667 \n",
353 | "Epoch 45/50\n",
354 | "30/30 [==============================] - 0s - loss: 0.5963 - acc: 0.8000 \n",
355 | "Epoch 46/50\n",
356 | "30/30 [==============================] - 0s - loss: 0.7851 - acc: 0.7000 \n",
357 | "Epoch 47/50\n",
358 | "30/30 [==============================] - 0s - loss: 0.5623 - acc: 0.8333 \n",
359 | "Epoch 48/50\n",
360 | "30/30 [==============================] - 0s - loss: 0.7574 - acc: 0.7333 \n",
361 | "Epoch 49/50\n",
362 | "30/30 [==============================] - 0s - loss: 0.4094 - acc: 0.8667 \n",
363 | "Epoch 50/50\n",
364 | "30/30 [==============================] - 0s - loss: 0.6300 - acc: 0.8000 \n",
365 | "Accuracy = 0.93\n"
366 | ]
367 | }
368 | ],
369 | "source": [
370 | "model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['accuracy'])\n",
371 | "# モデル訓練\n",
372 | "model.fit(test_X, test_Y, epochs=50, batch_size=1, verbose=1)\n",
373 | "\n",
374 | "loss, accuracy = model.evaluate(test_X, test_Y, verbose=0)\n",
375 | "print(\"Accuracy = {:.2f}\".format(accuracy))"
376 | ]
377 | },
378 | {
379 | "cell_type": "code",
380 | "execution_count": null,
381 | "metadata": {
382 | "collapsed": true
383 | },
384 | "outputs": [],
385 | "source": []
386 | }
387 | ],
388 | "metadata": {
389 | "kernelspec": {
390 | "display_name": "Python 3",
391 | "language": "python",
392 | "name": "python3"
393 | },
394 | "language_info": {
395 | "codemirror_mode": {
396 | "name": "ipython",
397 | "version": 3
398 | },
399 | "file_extension": ".py",
400 | "mimetype": "text/x-python",
401 | "name": "python",
402 | "nbconvert_exporter": "python",
403 | "pygments_lexer": "ipython3",
404 | "version": "3.5.0"
405 | }
406 | },
407 | "nbformat": 4,
408 | "nbformat_minor": 2
409 | }
410 |
--------------------------------------------------------------------------------