4 |
4 | | ⚠️ | 6 |
7 | Both questions and answers here are given by the community. Be careful and double check the answers before using them. 8 | If you see an error, please create a PR with a fix 9 | |
10 |
| ⚠️ | 6 |
7 | The answers here are given by the community. Be careful and double check the answers before using them. 8 | If you see an error, please create a PR with a fix 9 | |
10 |
34 |
35 |
36 | Write SQL queries to extract the following information:
37 |
38 | **1)** The number of active ads.
39 |
40 | ```sql
41 | SELECT count(*) FROM Ads WHERE status = 'active';
42 | ```
43 |
44 | 
70 |
71 | ```sql
72 | SELECT e.ad_id, e.event_type, count(*) as "count"
73 | FROM Events AS e
74 | GROUP BY e.ad_id, e.event_type
75 | ORDER BY e.ad_id, "count" DESC;
76 | ```
77 |
78 | 
83 |
84 | ```sql
85 | SELECT a.ad_id, e.event_type, e.date, count(*) as "count"
86 | FROM Ads AS a
87 | JOIN Events AS e
88 | ON a.ad_id = e.ad_id
89 | WHERE a.status = 'active'
90 | AND e.date >= DATEADD(week, -1, GETDATE())
91 | GROUP BY a.ad_id, e.event_type, e.date
92 | ORDER BY e.date ASC, "count" DESC;
93 | ```
94 |
95 | 
100 |
101 |
102 | ```sql
103 | SELECT a.campaign_id, e.event_type, count(*) as count
104 | FROM Ads AS a
105 | INNER JOIN Events AS e
106 | ON a.ad_id = e.ad_id
107 | GROUP BY a.campaign_id, e.event_type
108 | ORDER BY a.campaign_id, "count" DESC
109 | ```
110 |
111 | 
116 |
117 | ```sql
118 | -- for Postgres
119 |
120 | SELECT a.campaign_id AS camp_id, e.event_type, e.date, count(*)
121 | FROM Ads AS a
122 | INNER JOIN Events AS e
123 | ON a.ad_id = e.ad_id
124 | WHERE e.date >= DATEADD(week, -1, GETDATE())
125 | GROUP BY a.campaign_id, e.event_type, e.date
126 | ORDER BY a.campaign_id, e.date DESC, "count" DESC;
127 | ```
128 |
129 | 
134 |
135 | ```sql
136 | -- for Postgres
137 |
138 | SELECT impressions_clicks_table.ad_id,
139 | (impressions_clicks_table.clicks * 100 / impressions_clicks_table.impressions)::FLOAT || '%' AS CTR
140 | FROM
141 | (
142 | SELECT e.ad_id,
143 | SUM(CASE e.event_type WHEN 'impression' THEN 1 ELSE 0 END) impressions,
144 | SUM(CASE e.event_type WHEN 'click' THEN 1 ELSE 0 END) clicks
145 | FROM Events AS e
146 | GROUP BY e.ad_id
147 | ) AS impressions_clicks_table
148 | ORDER BY impressions_clicks_table.ad_id;
149 | ```
150 |
151 | 
156 |
157 | ```sql
158 | -- for Postgres
159 |
160 | SELECT conversions_clicks_table.ad_id,
161 | (conversions_clicks_table.conversions * 100 / conversions_clicks_table.clicks)::FLOAT || '%' AS CVR
162 | FROM
163 | (
164 | SELECT e.ad_id,
165 | SUM(CASE e.event_type WHEN 'conversion' THEN 1 ELSE 0 END) conversions,
166 | SUM(CASE e.event_type WHEN 'click' THEN 1 ELSE 0 END) clicks
167 | FROM Events AS e
168 | GROUP BY e.ad_id
169 | ) AS conversions_clicks_table
170 | ORDER BY conversions_clicks_table.ad_id;
171 | ```
172 |
173 | 
178 |
179 |
180 | ```sql
181 | -- for Postgres
182 |
183 | SELECT conversions_clicks_table.ad_id,
184 | conversions_clicks_table.date,
185 | conversions_clicks_table.hour,
186 | (impressions_clicks_table.clicks * 100 / impressions_clicks_table.impressions)::FLOAT || '%' AS CTR,
187 | (conversions_clicks_table.conversions * 100 / conversions_clicks_table.clicks)::FLOAT || '%' AS CVR
188 | FROM
189 | (
190 | SELECT e.ad_id, e.date, e.hour,
191 | SUM(CASE e.event_type WHEN 'conversion' THEN 1 ELSE 0 END) conversions,
192 | SUM(CASE e.event_type WHEN 'click' THEN 1 ELSE 0 END) clicks,
193 | SUM(CASE e.event_type WHEN 'impression' THEN 1 ELSE 0 END) impressions
194 | FROM Events AS e
195 | GROUP BY e.ad_id, e.date, e.hour
196 | ) AS conversions_clicks_table
197 | ORDER BY conversions_clicks_table.ad_id, conversions_clicks_table.date DESC, conversions_clicks_table.hour DESC, "CTR" DESC, "CVR" DESC;
198 | ```
199 |
200 | 
205 |
206 |
207 | ```sql
208 | -- for Postgres
209 |
210 | SELECT conversions_clicks_table.ad_id,
211 | conversions_clicks_table.date,
212 | conversions_clicks_table.source,
213 | (impressions_clicks_table.clicks * 100 / impressions_clicks_table.impressions)::FLOAT || '%' AS CTR
214 | FROM
215 | (
216 | SELECT e.ad_id, e.date, e.source,
217 | SUM(CASE e.event_type WHEN 'click' THEN 1 ELSE 0 END) clicks,
218 | SUM(CASE e.event_type WHEN 'impression' THEN 1 ELSE 0 END) impressions
219 | FROM Events AS e
220 | GROUP BY e.ad_id, e.date, e.source
221 | ) AS conversions_clicks_table
222 | ORDER BY conversions_clicks_table.ad_id, conversions_clicks_table.date DESC, conversions_clicks_table.source, "CTR" DESC;
223 | ```
224 |
225 | 
285 |
286 | ```python
287 | def mean(numbers):
288 | if len(numbers) > 0:
289 | return sum(numbers) / len(numbers)
290 | return float('NaN')
291 | ```
292 |
293 | 
302 |
303 | ```python
304 | from math import sqrt
305 | from statistics import mean
306 |
307 | def std_dev(numbers):
308 | if len(numbers) > 1:
309 | avg = mean(numbers)
310 | var = sum((i - avg) ** 2 for i in numbers) / (len(numbers) - 1)
311 | std = sqrt(var)
312 | return std
313 | return float('NaN')
314 | ```
315 |
316 | 
324 |
325 | ```python
326 | import math
327 |
328 | def rmse(y_true, y_pred):
329 | assert len(y_true) == len(y_pred), 'different sizes of the arguments'
330 | squares = sum((x - y)**2 for x, y in zip(y_true, y_pred))
331 | return math.sqrt(squares / len(y_true))
332 | ```
333 |
334 | 
413 |
414 | ```python
415 | def counter(lst):
416 | ans = {}
417 | for i in lst:
418 | site = i[-2:]
419 | ans[site] = ans.get(site, 0) + 1
420 | return ans
421 | ```
422 |
423 | 
428 |
429 | ```python
430 | def top_counter(lst):
431 | site_dict = counter(lst) # using last problem's solution
432 | top_keys = sorted(site_dict, reverse=True, key=site_dict.get)[:3]
433 | return {key: site_dict[key] for key in top_keys}
434 | ```
435 |
436 | 
474 |
475 | ```python
476 | def jaccard(a, b):
477 | return len(a & b) / len(a | b)
478 | ```
479 |
480 | 
487 |
488 | Where:
489 |
490 | * t is the token,
491 | * n(t) is the number of documents that t occurs in,
492 | * N is the total number of documents
493 |
494 | ```python
495 | from math import log10
496 |
497 | def idf1(docs):
498 | docs = [set(doc) for doc in docs]
499 | n_tokens = {}
500 | for doc in docs:
501 | for token in doc:
502 | n_tokens[token] = n_tokens.get(token, 0) + 1
503 | ans = {}
504 | for token in n_tokens:
505 | ans[token] = log10(len(docs) / (1 + n_tokens[token]))
506 | return ans
507 | ```
508 |
509 | ```python
510 | import math
511 |
512 | def idf2(docs):
513 | n_docs = len(docs)
514 |
515 | docs = [set(doc) for doc in docs]
516 | all_tokens = set.union(*docs)
517 |
518 | idf_coefficients = {}
519 | for token in all_tokens:
520 | n_docs_w_token = sum(token in doc for doc in docs)
521 | idf_c = math.log10(n_docs / (1 + n_docs_w_token))
522 | idf_coefficients[token] = idf_c
523 |
524 | return idf_coefficients
525 | ```
526 |
527 | 
536 |
537 | The higher the PMI, the more likely these two tokens form a collection. We can estimate PMI by counting:
538 |
539 | 
540 |
541 | Where:
542 | * N is the total number of tokens in the text,
543 | * c(t1, t2) is the number of times t1 and t2 appear together,
544 | * c(t1) and c(t2) — the number of times they appear separately.
545 |
546 | ```python
547 | import math
548 |
549 | def pmi(docs):
550 | n_pairs = {}
551 | n_tokens = {}
552 |
553 | docs = [tuple(doc) for doc in docs]
554 | for doc in docs:
555 | for token in doc:
556 | n_tokens[token] = n_tokens.get(token, 0) + 1
557 |
558 | n_pairs[doc] = n_pairs.get(doc, 0) + 1
559 | ans = {}
560 | for pair in n_pairs:
561 | ans[pair] = math.log2(n_pairs[pair] * (sum(n_tokens.values())) / (n_tokens[pair[0]] * n_tokens[pair[1]]))
562 |
563 | srt = sorted(ans.items(), key=lambda x: x[1], reverse=True)[:10]
564 |
565 | return srt
566 | ```
567 |
568 |
569 | 
751 |
752 | ```python
753 | def flip_bt(head):
754 | if head is not None:
755 | head.left, head.right = head.right, head.left
756 | flip_bt(head.left)
757 | flip_bt(head.right)
758 | ```
759 |
760 | | ⚠️ | 6 |
7 | The answers here are given by the community. Be careful and double check the answers before using them. 8 | If you see an error, please create a PR with a fix 9 | |
10 |