└── ABtest.ipynb


/ABtest.ipynb:
--------------------------------------------------------------------------------
  1 | {
  2 |  "cells": [
  3 |   {
  4 |    "cell_type": "code",
  5 |    "execution_count": 2,
  6 |    "metadata": {},
  7 |    "outputs": [],
  8 |    "source": [
  9 |     "import pandas as pd\n",
 10 |     "import numpy as np\n",
 11 |     "import matplotlib.pyplot as plt\n",
 12 |     "%matplotlib inline"
 13 |    ]
 14 |   },
 15 |   {
 16 |    "cell_type": "markdown",
 17 |    "metadata": {},
 18 |    "source": [
 19 |     "#### 数据集解释：\n",
 20 |     "#### emb_tb_2.csv:用户特征数据集\n",
 21 |     "#### effct_tb.csv:广告点击情况数据集\n",
 22 |     "#### seed_cand_tb.csv:用户类型数据集\n",
 23 |     "#### 本报告主要使用广告点击情况数据，主要涉及字段如下：\n",
 24 |     "#### dmp_id:营销策略编号（源数据文档未做说明，这个根据情况设定为1.对照组，2.营销策略一，3.营销策略二）\n",
 25 |     "#### user_id:支付宝用户id\n",
 26 |     "#### label:用户当天是否点击活动广告(0:未点击，1：点击)"
 27 |    ]
 28 |   },
 29 |   {
 30 |    "cell_type": "code",
 31 |    "execution_count": 3,
 32 |    "metadata": {},
 33 |    "outputs": [
 34 |     {
 35 |      "data": {
 36 |       "text/html": [
 37 |        "<div>\n",
 38 |        "<style scoped>\n",
 39 |        "    .dataframe tbody tr th:only-of-type {\n",
 40 |        "        vertical-align: middle;\n",
 41 |        "    }\n",
 42 |        "\n",
 43 |        "    .dataframe tbody tr th {\n",
 44 |        "        vertical-align: top;\n",
 45 |        "    }\n",
 46 |        "\n",
 47 |        "    .dataframe thead th {\n",
 48 |        "        text-align: right;\n",
 49 |        "    }\n",
 50 |        "</style>\n",
 51 |        "<table border=\"1\" class=\"dataframe\">\n",
 52 |        "  <thead>\n",
 53 |        "    <tr style=\"text-align: right;\">\n",
 54 |        "      <th></th>\n",
 55 |        "      <th>user_id</th>\n",
 56 |        "      <th>label</th>\n",
 57 |        "      <th>dmp_id</th>\n",
 58 |        "    </tr>\n",
 59 |        "  </thead>\n",
 60 |        "  <tbody>\n",
 61 |        "    <tr>\n",
 62 |        "      <th>0</th>\n",
 63 |        "      <td>1</td>\n",
 64 |        "      <td>0</td>\n",
 65 |        "      <td>1</td>\n",
 66 |        "    </tr>\n",
 67 |        "    <tr>\n",
 68 |        "      <th>1</th>\n",
 69 |        "      <td>1000004</td>\n",
 70 |        "      <td>0</td>\n",
 71 |        "      <td>1</td>\n",
 72 |        "    </tr>\n",
 73 |        "    <tr>\n",
 74 |        "      <th>2</th>\n",
 75 |        "      <td>1000004</td>\n",
 76 |        "      <td>0</td>\n",
 77 |        "      <td>2</td>\n",
 78 |        "    </tr>\n",
 79 |        "  </tbody>\n",
 80 |        "</table>\n",
 81 |        "</div>"
 82 |       ],
 83 |       "text/plain": [
 84 |        "   user_id  label  dmp_id\n",
 85 |        "0        1      0       1\n",
 86 |        "1  1000004      0       1\n",
 87 |        "2  1000004      0       2"
 88 |       ]
 89 |      },
 90 |      "execution_count": 3,
 91 |      "metadata": {},
 92 |      "output_type": "execute_result"
 93 |     }
 94 |    ],
 95 |    "source": [
 96 |     "#加载数据\n",
 97 |     "data = pd.read_csv('F:/数据分析/ABtest_data/effect_tb.csv',header = None)\n",
 98 |     "data.columns = [\"dt\",\"user_id\",\"label\",\"dmp_id\"]\n",
 99 |     "#日志天数属性用户上，删除该列\n",
100 |     "data = data.drop(columns = \"dt\")\n",
101 |     "data.head(3)"
102 |    ]
103 |   },
104 |   {
105 |    "cell_type": "code",
106 |    "execution_count": 4,
107 |    "metadata": {},
108 |    "outputs": [
109 |     {
110 |      "data": {
111 |       "text/html": [
112 |        "<div>\n",
113 |        "<style scoped>\n",
114 |        "    .dataframe tbody tr th:only-of-type {\n",
115 |        "        vertical-align: middle;\n",
116 |        "    }\n",
117 |        "\n",
118 |        "    .dataframe tbody tr th {\n",
119 |        "        vertical-align: top;\n",
120 |        "    }\n",
121 |        "\n",
122 |        "    .dataframe thead th {\n",
123 |        "        text-align: right;\n",
124 |        "    }\n",
125 |        "</style>\n",
126 |        "<table border=\"1\" class=\"dataframe\">\n",
127 |        "  <thead>\n",
128 |        "    <tr style=\"text-align: right;\">\n",
129 |        "      <th></th>\n",
130 |        "      <th>user_id</th>\n",
131 |        "      <th>label</th>\n",
132 |        "      <th>dmp_id</th>\n",
133 |        "    </tr>\n",
134 |        "  </thead>\n",
135 |        "  <tbody>\n",
136 |        "    <tr>\n",
137 |        "      <th>count</th>\n",
138 |        "      <td>2.645958e+06</td>\n",
139 |        "      <td>2.645958e+06</td>\n",
140 |        "      <td>2.645958e+06</td>\n",
141 |        "    </tr>\n",
142 |        "    <tr>\n",
143 |        "      <th>mean</th>\n",
144 |        "      <td>3.112995e+06</td>\n",
145 |        "      <td>1.456297e-02</td>\n",
146 |        "      <td>1.395761e+00</td>\n",
147 |        "    </tr>\n",
148 |        "    <tr>\n",
149 |        "      <th>std</th>\n",
150 |        "      <td>1.828262e+06</td>\n",
151 |        "      <td>1.197952e-01</td>\n",
152 |        "      <td>6.920480e-01</td>\n",
153 |        "    </tr>\n",
154 |        "    <tr>\n",
155 |        "      <th>min</th>\n",
156 |        "      <td>1.000000e+00</td>\n",
157 |        "      <td>0.000000e+00</td>\n",
158 |        "      <td>1.000000e+00</td>\n",
159 |        "    </tr>\n",
160 |        "    <tr>\n",
161 |        "      <th>25%</th>\n",
162 |        "      <td>1.526772e+06</td>\n",
163 |        "      <td>0.000000e+00</td>\n",
164 |        "      <td>1.000000e+00</td>\n",
165 |        "    </tr>\n",
166 |        "    <tr>\n",
167 |        "      <th>50%</th>\n",
168 |        "      <td>3.062184e+06</td>\n",
169 |        "      <td>0.000000e+00</td>\n",
170 |        "      <td>1.000000e+00</td>\n",
171 |        "    </tr>\n",
172 |        "    <tr>\n",
173 |        "      <th>75%</th>\n",
174 |        "      <td>4.721132e+06</td>\n",
175 |        "      <td>0.000000e+00</td>\n",
176 |        "      <td>2.000000e+00</td>\n",
177 |        "    </tr>\n",
178 |        "    <tr>\n",
179 |        "      <th>max</th>\n",
180 |        "      <td>6.265402e+06</td>\n",
181 |        "      <td>1.000000e+00</td>\n",
182 |        "      <td>3.000000e+00</td>\n",
183 |        "    </tr>\n",
184 |        "  </tbody>\n",
185 |        "</table>\n",
186 |        "</div>"
187 |       ],
188 |       "text/plain": [
189 |        "            user_id         label        dmp_id\n",
190 |        "count  2.645958e+06  2.645958e+06  2.645958e+06\n",
191 |        "mean   3.112995e+06  1.456297e-02  1.395761e+00\n",
192 |        "std    1.828262e+06  1.197952e-01  6.920480e-01\n",
193 |        "min    1.000000e+00  0.000000e+00  1.000000e+00\n",
194 |        "25%    1.526772e+06  0.000000e+00  1.000000e+00\n",
195 |        "50%    3.062184e+06  0.000000e+00  1.000000e+00\n",
196 |        "75%    4.721132e+06  0.000000e+00  2.000000e+00\n",
197 |        "max    6.265402e+06  1.000000e+00  3.000000e+00"
198 |       ]
199 |      },
200 |      "execution_count": 4,
201 |      "metadata": {},
202 |      "output_type": "execute_result"
203 |     }
204 |    ],
205 |    "source": [
206 |     "# 查看计数，平均数，方差，最小值和四分位数，最大值\n",
207 |     "data.describe()"
208 |    ]
209 |   },
210 |   {
211 |    "cell_type": "markdown",
212 |    "metadata": {},
213 |    "source": [
214 |     "### 重复值处理"
215 |    ]
216 |   },
217 |   {
218 |    "cell_type": "code",
219 |    "execution_count": 5,
220 |    "metadata": {},
221 |    "outputs": [
222 |     {
223 |      "data": {
224 |       "text/plain": [
225 |        "(2645958, 3)"
226 |       ]
227 |      },
228 |      "execution_count": 5,
229 |      "metadata": {},
230 |      "output_type": "execute_result"
231 |     }
232 |    ],
233 |    "source": [
234 |     "data.shape"
235 |    ]
236 |   },
237 |   {
238 |    "cell_type": "code",
239 |    "execution_count": 6,
240 |    "metadata": {},
241 |    "outputs": [
242 |     {
243 |      "data": {
244 |       "text/plain": [
245 |        "user_id    2410683\n",
246 |        "label            2\n",
247 |        "dmp_id           3\n",
248 |        "dtype: int64"
249 |       ]
250 |      },
251 |      "execution_count": 6,
252 |      "metadata": {},
253 |      "output_type": "execute_result"
254 |     }
255 |    ],
256 |    "source": [
257 |     "data.nunique()"
258 |    ]
259 |   },
260 |   {
261 |    "cell_type": "markdown",
262 |    "metadata": {},
263 |    "source": [
264 |     "数据行与独立用户数不统一，检查是否存在重复行。"
265 |    ]
266 |   },
267 |   {
268 |    "cell_type": "code",
269 |    "execution_count": 7,
270 |    "metadata": {},
271 |    "outputs": [
272 |     {
273 |      "data": {
274 |       "text/html": [
275 |        "<div>\n",
276 |        "<style scoped>\n",
277 |        "    .dataframe tbody tr th:only-of-type {\n",
278 |        "        vertical-align: middle;\n",
279 |        "    }\n",
280 |        "\n",
281 |        "    .dataframe tbody tr th {\n",
282 |        "        vertical-align: top;\n",
283 |        "    }\n",
284 |        "\n",
285 |        "    .dataframe thead th {\n",
286 |        "        text-align: right;\n",
287 |        "    }\n",
288 |        "</style>\n",
289 |        "<table border=\"1\" class=\"dataframe\">\n",
290 |        "  <thead>\n",
291 |        "    <tr style=\"text-align: right;\">\n",
292 |        "      <th></th>\n",
293 |        "      <th>user_id</th>\n",
294 |        "      <th>label</th>\n",
295 |        "      <th>dmp_id</th>\n",
296 |        "    </tr>\n",
297 |        "  </thead>\n",
298 |        "  <tbody>\n",
299 |        "    <tr>\n",
300 |        "      <th>8529</th>\n",
301 |        "      <td>1027</td>\n",
302 |        "      <td>0</td>\n",
303 |        "      <td>1</td>\n",
304 |        "    </tr>\n",
305 |        "    <tr>\n",
306 |        "      <th>1485546</th>\n",
307 |        "      <td>1027</td>\n",
308 |        "      <td>0</td>\n",
309 |        "      <td>1</td>\n",
310 |        "    </tr>\n",
311 |        "    <tr>\n",
312 |        "      <th>1579415</th>\n",
313 |        "      <td>1471</td>\n",
314 |        "      <td>0</td>\n",
315 |        "      <td>1</td>\n",
316 |        "    </tr>\n",
317 |        "    <tr>\n",
318 |        "      <th>127827</th>\n",
319 |        "      <td>1471</td>\n",
320 |        "      <td>0</td>\n",
321 |        "      <td>1</td>\n",
322 |        "    </tr>\n",
323 |        "    <tr>\n",
324 |        "      <th>404862</th>\n",
325 |        "      <td>2468</td>\n",
326 |        "      <td>0</td>\n",
327 |        "      <td>1</td>\n",
328 |        "    </tr>\n",
329 |        "    <tr>\n",
330 |        "      <th>...</th>\n",
331 |        "      <td>...</td>\n",
332 |        "      <td>...</td>\n",
333 |        "      <td>...</td>\n",
334 |        "    </tr>\n",
335 |        "    <tr>\n",
336 |        "      <th>1382121</th>\n",
337 |        "      <td>6264633</td>\n",
338 |        "      <td>0</td>\n",
339 |        "      <td>1</td>\n",
340 |        "    </tr>\n",
341 |        "    <tr>\n",
342 |        "      <th>1382245</th>\n",
343 |        "      <td>6264940</td>\n",
344 |        "      <td>0</td>\n",
345 |        "      <td>1</td>\n",
346 |        "    </tr>\n",
347 |        "    <tr>\n",
348 |        "      <th>2575140</th>\n",
349 |        "      <td>6264940</td>\n",
350 |        "      <td>0</td>\n",
351 |        "      <td>1</td>\n",
352 |        "    </tr>\n",
353 |        "    <tr>\n",
354 |        "      <th>1382306</th>\n",
355 |        "      <td>6265082</td>\n",
356 |        "      <td>0</td>\n",
357 |        "      <td>3</td>\n",
358 |        "    </tr>\n",
359 |        "    <tr>\n",
360 |        "      <th>2575171</th>\n",
361 |        "      <td>6265082</td>\n",
362 |        "      <td>0</td>\n",
363 |        "      <td>3</td>\n",
364 |        "    </tr>\n",
365 |        "  </tbody>\n",
366 |        "</table>\n",
367 |        "<p>25966 rows × 3 columns</p>\n",
368 |        "</div>"
369 |       ],
370 |       "text/plain": [
371 |        "         user_id  label  dmp_id\n",
372 |        "8529        1027      0       1\n",
373 |        "1485546     1027      0       1\n",
374 |        "1579415     1471      0       1\n",
375 |        "127827      1471      0       1\n",
376 |        "404862      2468      0       1\n",
377 |        "...          ...    ...     ...\n",
378 |        "1382121  6264633      0       1\n",
379 |        "1382245  6264940      0       1\n",
380 |        "2575140  6264940      0       1\n",
381 |        "1382306  6265082      0       3\n",
382 |        "2575171  6265082      0       3\n",
383 |        "\n",
384 |        "[25966 rows x 3 columns]"
385 |       ]
386 |      },
387 |      "execution_count": 7,
388 |      "metadata": {},
389 |      "output_type": "execute_result"
390 |     }
391 |    ],
392 |    "source": [
393 |     "data[data.duplicated(keep = False)].sort_values(by=[\"user_id\"])"
394 |    ]
395 |   },
396 |   {
397 |    "cell_type": "code",
398 |    "execution_count": 8,
399 |    "metadata": {},
400 |    "outputs": [
401 |     {
402 |      "data": {
403 |       "text/html": [
404 |        "<div>\n",
405 |        "<style scoped>\n",
406 |        "    .dataframe tbody tr th:only-of-type {\n",
407 |        "        vertical-align: middle;\n",
408 |        "    }\n",
409 |        "\n",
410 |        "    .dataframe tbody tr th {\n",
411 |        "        vertical-align: top;\n",
412 |        "    }\n",
413 |        "\n",
414 |        "    .dataframe thead th {\n",
415 |        "        text-align: right;\n",
416 |        "    }\n",
417 |        "</style>\n",
418 |        "<table border=\"1\" class=\"dataframe\">\n",
419 |        "  <thead>\n",
420 |        "    <tr style=\"text-align: right;\">\n",
421 |        "      <th></th>\n",
422 |        "      <th>user_id</th>\n",
423 |        "      <th>label</th>\n",
424 |        "      <th>dmp_id</th>\n",
425 |        "    </tr>\n",
426 |        "  </thead>\n",
427 |        "  <tbody>\n",
428 |        "  </tbody>\n",
429 |        "</table>\n",
430 |        "</div>"
431 |       ],
432 |       "text/plain": [
433 |        "Empty DataFrame\n",
434 |        "Columns: [user_id, label, dmp_id]\n",
435 |        "Index: []"
436 |       ]
437 |      },
438 |      "execution_count": 8,
439 |      "metadata": {},
440 |      "output_type": "execute_result"
441 |     }
442 |    ],
443 |    "source": [
444 |     "#删除重复值\n",
445 |     "data = data.drop_duplicates()\n",
446 |     "data[data.duplicated(keep = False)]"
447 |    ]
448 |   },
449 |   {
450 |    "cell_type": "markdown",
451 |    "metadata": {},
452 |    "source": [
453 |     "### 3.空值处理"
454 |    ]
455 |   },
456 |   {
457 |    "cell_type": "code",
458 |    "execution_count": 9,
459 |    "metadata": {},
460 |    "outputs": [
461 |     {
462 |      "name": "stdout",
463 |      "output_type": "stream",
464 |      "text": [
465 |       "<class 'pandas.core.frame.DataFrame'>\n",
466 |       "Int64Index: 2632975 entries, 0 to 2645957\n",
467 |       "Data columns (total 3 columns):\n",
468 |       " #   Column   Non-Null Count    Dtype\n",
469 |       "---  ------   --------------    -----\n",
470 |       " 0   user_id  2632975 non-null  int64\n",
471 |       " 1   label    2632975 non-null  int64\n",
472 |       " 2   dmp_id   2632975 non-null  int64\n",
473 |       "dtypes: int64(3)\n",
474 |       "memory usage: 80.4 MB\n"
475 |      ]
476 |     }
477 |    ],
478 |    "source": [
479 |     "data.info(null_counts = True)"
480 |    ]
481 |   },
482 |   {
483 |    "cell_type": "code",
484 |    "execution_count": 10,
485 |    "metadata": {},
486 |    "outputs": [],
487 |    "source": [
488 |     "#### 数据集没有空值可以不用处理"
489 |    ]
490 |   },
491 |   {
492 |    "cell_type": "markdown",
493 |    "metadata": {},
494 |    "source": [
495 |     "### 4.异常值检查\n",
496 |     "##### 通过透视表格检查各属性字段是否存在不合理取值"
497 |    ]
498 |   },
499 |   {
500 |    "cell_type": "code",
501 |    "execution_count": 11,
502 |    "metadata": {},
503 |    "outputs": [
504 |     {
505 |      "data": {
506 |       "text/html": [
507 |        "<div>\n",
508 |        "<style scoped>\n",
509 |        "    .dataframe tbody tr th:only-of-type {\n",
510 |        "        vertical-align: middle;\n",
511 |        "    }\n",
512 |        "\n",
513 |        "    .dataframe tbody tr th {\n",
514 |        "        vertical-align: top;\n",
515 |        "    }\n",
516 |        "\n",
517 |        "    .dataframe thead th {\n",
518 |        "        text-align: right;\n",
519 |        "    }\n",
520 |        "</style>\n",
521 |        "<table border=\"1\" class=\"dataframe\">\n",
522 |        "  <thead>\n",
523 |        "    <tr style=\"text-align: right;\">\n",
524 |        "      <th>label</th>\n",
525 |        "      <th>0</th>\n",
526 |        "      <th>1</th>\n",
527 |        "      <th>All</th>\n",
528 |        "    </tr>\n",
529 |        "    <tr>\n",
530 |        "      <th>dmp_id</th>\n",
531 |        "      <th></th>\n",
532 |        "      <th></th>\n",
533 |        "      <th></th>\n",
534 |        "    </tr>\n",
535 |        "  </thead>\n",
536 |        "  <tbody>\n",
537 |        "    <tr>\n",
538 |        "      <th>1</th>\n",
539 |        "      <td>1881745</td>\n",
540 |        "      <td>23918</td>\n",
541 |        "      <td>1905663</td>\n",
542 |        "    </tr>\n",
543 |        "    <tr>\n",
544 |        "      <th>2</th>\n",
545 |        "      <td>404811</td>\n",
546 |        "      <td>6296</td>\n",
547 |        "      <td>411107</td>\n",
548 |        "    </tr>\n",
549 |        "    <tr>\n",
550 |        "      <th>3</th>\n",
551 |        "      <td>307923</td>\n",
552 |        "      <td>8282</td>\n",
553 |        "      <td>316205</td>\n",
554 |        "    </tr>\n",
555 |        "    <tr>\n",
556 |        "      <th>All</th>\n",
557 |        "      <td>2594479</td>\n",
558 |        "      <td>38496</td>\n",
559 |        "      <td>2632975</td>\n",
560 |        "    </tr>\n",
561 |        "  </tbody>\n",
562 |        "</table>\n",
563 |        "</div>"
564 |       ],
565 |       "text/plain": [
566 |        "label         0      1      All\n",
567 |        "dmp_id                         \n",
568 |        "1       1881745  23918  1905663\n",
569 |        "2        404811   6296   411107\n",
570 |        "3        307923   8282   316205\n",
571 |        "All     2594479  38496  2632975"
572 |       ]
573 |      },
574 |      "execution_count": 11,
575 |      "metadata": {},
576 |      "output_type": "execute_result"
577 |     }
578 |    ],
579 |    "source": [
580 |     "data.pivot_table(index = \"dmp_id\",columns = \"label\",values = \"user_id\",aggfunc = \"count\",margins = True)"
581 |    ]
582 |   },
583 |   {
584 |    "cell_type": "markdown",
585 |    "metadata": {},
586 |    "source": [
587 |     "属性字段没有发现一场自，无需进行处理"
588 |    ]
589 |   },
590 |   {
591 |    "cell_type": "markdown",
592 |    "metadata": {},
593 |    "source": [
594 |     "### 5.数据类型"
595 |    ]
596 |   },
597 |   {
598 |    "cell_type": "code",
599 |    "execution_count": 12,
600 |    "metadata": {},
601 |    "outputs": [
602 |     {
603 |      "data": {
604 |       "text/plain": [
605 |        "user_id    int64\n",
606 |        "label      int64\n",
607 |        "dmp_id     int64\n",
608 |        "dtype: object"
609 |       ]
610 |      },
611 |      "execution_count": 12,
612 |      "metadata": {},
613 |      "output_type": "execute_result"
614 |     }
615 |    ],
616 |    "source": [
617 |     "data.dtypes"
618 |    ]
619 |   },
620 |   {
621 |    "cell_type": "markdown",
622 |    "metadata": {},
623 |    "source": [
624 |     "数据类型正常不需要转换"
625 |    ]
626 |   },
627 |   {
628 |    "cell_type": "markdown",
629 |    "metadata": {},
630 |    "source": [
631 |     "### 2.2样本容量检验\n",
632 |     "一般采用样本量计算工具：sample size calculator.使用可以看自己写的ABtest学习文档"
633 |    ]
634 |   },
635 |   {
636 |    "cell_type": "code",
637 |    "execution_count": 13,
638 |    "metadata": {},
639 |    "outputs": [
640 |     {
641 |      "data": {
642 |       "text/plain": [
643 |        "0.012551012429794775"
644 |       ]
645 |      },
646 |      "execution_count": 13,
647 |      "metadata": {},
648 |      "output_type": "execute_result"
649 |     }
650 |    ],
651 |    "source": [
652 |     "data[data[\"dmp_id\"] == 1][\"label\"].mean()"
653 |    ]
654 |   },
655 |   {
656 |    "cell_type": "markdown",
657 |    "metadata": {},
658 |    "source": [
659 |     "这个数字是对照组的点击率，为1.26%，我们需要的新的营销策略能让广告点击率至少提高1个百分点，那么\n",
660 |     "我们在网站中https://www.evanmiller.org/ab-testing/sample-size.html\n",
661 |     "    baseline conversion rate 框中输入1.26\n",
662 |     "    minimum detectable effect 框中输入 1\n",
663 |     "    计算得出我们实验所需要的的最小样本量为2167"
664 |    ]
665 |   },
666 |   {
667 |    "cell_type": "code",
668 |    "execution_count": 14,
669 |    "metadata": {},
670 |    "outputs": [
671 |     {
672 |      "data": {
673 |       "text/plain": [
674 |        "1    1905663\n",
675 |        "2     411107\n",
676 |        "3     316205\n",
677 |        "Name: dmp_id, dtype: int64"
678 |       ]
679 |      },
680 |      "execution_count": 14,
681 |      "metadata": {},
682 |      "output_type": "execute_result"
683 |     }
684 |    ],
685 |    "source": [
686 |     "#查看每个组中的样本数\n",
687 |     "data[\"dmp_id\"].value_counts()"
688 |    ]
689 |   },
690 |   {
691 |    "cell_type": "markdown",
692 |    "metadata": {},
693 |    "source": [
694 |     "可以看到三种策略的样本数都比我们最小样本数量大，因此样本合适"
695 |    ]
696 |   },
697 |   {
698 |    "cell_type": "code",
699 |    "execution_count": 15,
700 |    "metadata": {},
701 |    "outputs": [],
702 |    "source": [
703 |     "#保存清洗好的数据备用\n",
704 |     "data.to_csv(\"F:/数据分析/ABtest_data/output.csv\",index = False)"
705 |    ]
706 |   },
707 |   {
708 |    "cell_type": "markdown",
709 |    "metadata": {},
710 |    "source": [
711 |     "### 3.假设验证\n",
712 |     "先观察几组实验的点击情况"
713 |    ]
714 |   },
715 |   {
716 |    "cell_type": "code",
717 |    "execution_count": 16,
718 |    "metadata": {},
719 |    "outputs": [
720 |     {
721 |      "name": "stdout",
722 |      "output_type": "stream",
723 |      "text": [
724 |       "对照组： 0.012551012429794775\n",
725 |       "策略一组： 0.015314747742072015\n",
726 |       "策略二组： 0.026191869198779274\n"
727 |      ]
728 |     }
729 |    ],
730 |    "source": [
731 |     "print(\"对照组：\",data[data[\"dmp_id\"] ==1][\"label\"].mean())\n",
732 |     "print(\"策略一组：\",data[data[\"dmp_id\"] ==2][\"label\"].mean())\n",
733 |     "print(\"策略二组：\",data[data[\"dmp_id\"] ==3][\"label\"].mean())"
734 |    ]
735 |   },
736 |   {
737 |    "cell_type": "markdown",
738 |    "metadata": {},
739 |    "source": [
740 |     "从点击率来看，策略一和策略二在对照组的基础上都有一定的提升。\n",
741 |     "其中策略一提高了0.2个百分点，策略二提高了1.3个百分点，只有策略二满足了我们对点击率提升最小值的要求\n",
742 |     "接下来需要进行假设验证，来看看策略二的点击率提升是否显著"
743 |    ]
744 |   },
745 |   {
746 |    "cell_type": "markdown",
747 |    "metadata": {},
748 |    "source": [
749 |     "#### a.零假设和责备假设\n",
750 |     "##### 记对照组点击率为p1，策略二点击率为p2，则：\n",
751 |     "零假设设为H0：p1>=p2\n",
752 |     "备择假设为H1：p1< p2\n",
753 |     "##### b.分布类型，检验类型和显著性水平\n",
754 |     "样本服从两点分布，独立双样本，样本大小为n>30，总体均值和标准差未知，所以采用z检验，显著性水平取0.05"
755 |    ]
756 |   },
757 |   {
758 |    "cell_type": "markdown",
759 |    "metadata": {},
760 |    "source": [
761 |     "#### 3.1公式计算"
762 |    ]
763 |   },
764 |   {
765 |    "cell_type": "code",
766 |    "execution_count": 20,
767 |    "metadata": {},
768 |    "outputs": [
769 |     {
770 |      "name": "stdout",
771 |      "output_type": "stream",
772 |      "text": [
773 |       "总和点击率： 0.014492310074225832\n"
774 |      ]
775 |     },
776 |     {
777 |      "name": "stderr",
778 |      "output_type": "stream",
779 |      "text": [
780 |       "C:\\Users\\18042\\anaconda3\\lib\\site-packages\\ipykernel_launcher.py:6: UserWarning: Boolean Series key will be reindexed to match DataFrame index.\n",
781 |       "  \n",
782 |       "C:\\Users\\18042\\anaconda3\\lib\\site-packages\\ipykernel_launcher.py:7: UserWarning: Boolean Series key will be reindexed to match DataFrame index.\n",
783 |       "  import sys\n"
784 |      ]
785 |     }
786 |    ],
787 |    "source": [
788 |     "#用户数\n",
789 |     "n_old = len(data[data.dmp_id == 1]) #对照组\n",
790 |     "n_new = len(data[data.dmp_id == 3]) #策略二\n",
791 |     "\n",
792 |     "#点击数\n",
793 |     "c_old = len(data[data.dmp_id == 1][data.label == 1])\n",
794 |     "c_new = len(data[data.dmp_id == 3][data.label == 1])\n",
795 |     "\n",
796 |     "#计算点击率\n",
797 |     "try:\n",
798 |     "    if c_new == 0:\n",
799 |     "        print(\"不做运算\")\n",
800 |     "    else:\n",
801 |     "        r_old = c_old / n_old\n",
802 |     "        r_new = c_new / c_new\n",
803 |     "except:\n",
804 |     "    print(\"除数为0\")\n",
805 |     "#总和点击率\n",
806 |     "r = (c_old + c_new) / (n_old + n_new)\n",
807 |     "print(\"总和点击率：\", r)"
808 |    ]
809 |   },
810 |   {
811 |    "cell_type": "code",
812 |    "execution_count": 22,
813 |    "metadata": {},
814 |    "outputs": [
815 |     {
816 |      "name": "stdout",
817 |      "output_type": "stream",
818 |      "text": [
819 |       "检验统计量z: -4302.928619508961\n"
820 |      ]
821 |     }
822 |    ],
823 |    "source": [
824 |     "#计算检验统计量Z\n",
825 |     "z = (r_old - r_new) / np.sqrt(r*(1-r)*(1/n_old + 1/n_new))\n",
826 |     "print(\"检验统计量z:\", z)"
827 |    ]
828 |   },
829 |   {
830 |    "cell_type": "code",
831 |    "execution_count": 23,
832 |    "metadata": {},
833 |    "outputs": [
834 |     {
835 |      "data": {
836 |       "text/plain": [
837 |        "-1.6448536269514729"
838 |       ]
839 |      },
840 |      "execution_count": 23,
841 |      "metadata": {},
842 |      "output_type": "execute_result"
843 |     }
844 |    ],
845 |    "source": [
846 |     "#看显著水平0.05对应的Z的分位数\n",
847 |     "from scipy.stats import norm\n",
848 |     "z_alpha = norm.ppf(0.05)\n",
849 |     "z_alpha"
850 |    ]
851 |   },
852 |   {
853 |    "cell_type": "markdown",
854 |    "metadata": {},
855 |    "source": [
856 |     "z_aloha = -1.644，检验统计量z为-4032 ,该检验为左侧单尾检验，拒绝域为{z<z_alpha}.所以我们可以得出结论，原假设不成立，策略二点击率的提升在统计上是显著的"
857 |    ]
858 |   },
859 |   {
860 |    "cell_type": "code",
861 |    "execution_count": 24,
862 |    "metadata": {},
863 |    "outputs": [
864 |     {
865 |      "name": "stdout",
866 |      "output_type": "stream",
867 |      "text": [
868 |       "检验统计量z： -59.44168632985996 ，p值： 0.0\n"
869 |      ]
870 |     }
871 |    ],
872 |    "source": [
873 |     "import statsmodels.stats.proportion as sp\n",
874 |     "z_score, p = sp.proportions_ztest([c_old, c_new],[n_old, n_new], alternative = \"smaller\")\n",
875 |     "print(\"检验统计量z：\",z_score,\"，p值：\", p)"
876 |    ]
877 |   },
878 |   {
879 |    "cell_type": "markdown",
880 |    "metadata": {},
881 |    "source": [
882 |     "p值约等于0.p<显著值，所以结论一样，拒绝原假设。\n",
883 |     "综上所述，两种营销策略中，策略二对广告点击率有显著提升效果，且相较于对照组点击率提升了近一倍，因而在两组营销策略中应选择第二组进行推广"
884 |    ]
885 |   },
886 |   {
887 |    "cell_type": "code",
888 |    "execution_count": null,
889 |    "metadata": {},
890 |    "outputs": [],
891 |    "source": []
892 |   }
893 |  ],
894 |  "metadata": {
895 |   "kernelspec": {
896 |    "display_name": "Python 3",
897 |    "language": "python",
898 |    "name": "python3"
899 |   },
900 |   "language_info": {
901 |    "codemirror_mode": {
902 |     "name": "ipython",
903 |     "version": 3
904 |    },
905 |    "file_extension": ".py",
906 |    "mimetype": "text/x-python",
907 |    "name": "python",
908 |    "nbconvert_exporter": "python",
909 |    "pygments_lexer": "ipython3",
910 |    "version": "3.7.6"
911 |   }
912 |  },
913 |  "nbformat": 4,
914 |  "nbformat_minor": 4
915 | }
916 | 


--------------------------------------------------------------------------------