「データ分析」の解像度を上げたい / wanna-understand-data-analysis

2020年1月16日 Data Gateway Talk vol.5 (https://data-gateway-talk.connpass.com/event/155457/) でのLT資料です。

「データ分析」って抽象的なモノの解像度を上げて理解することで，「データ分析」に求める期待値のズレをなくしたいよねって話です。

1. ©2020 Wantedly, Inc. ʮσʔλ෼ੳʯͷղ૾౓Λ্͍͛ͨ ೥຤೥࢝ɼʮσʔλ෼ੳʯʹࢥ͍Λ஘ͤͯΈͨɻ Data Gateway Talk vol.5 16.Jan.2020 - Yuya Matsumura @yu-ya4
2. ©2020 Wantedly, Inc. ✓ Yuya Matsumura ✓ Wantedly, Inc. Recommendation Team ✓ Data Science, Team Lead ✓ Interested in Information Retrieval, Machine Learning Self-Introduction @yu-ya4 @yu__ya4
3. ©2020 Wantedly, Inc. ʮσʔλ෼ੳʯ͏·͍͍ͬͯ͘·͔͢ʁ
4. ©2020 Wantedly, Inc. ͜Μͳܦݧ͋Γ·ͤΜ͔ʁ
5. ©2020 Wantedly, Inc. • ৽ઃͷʮσʔλ෼ੳʯνʔϜʹϏδωενʔϜ͔Βͷ SQL ࡞੒ґཔ͕ࡴ౸͠ ͯຊདྷߦ͍͍ͨʮσʔλ෼ੳʯ͕ߦ͑ͳ͍ɻ • ʮઌ݄ͷചΓ্͛͸͍͘Β͔ʁʯͱฉ͔ΕͯʮʓʓສԁͰ͢ɻʯͱճ౴ͨ͠ ΒɼʮͰɼ݁ہͦΕͬͯͲ͏ͳͷ…ʁʯͱϚωʔδϟʹݴΘΕͯ͠·ͬͨɻ • Kaggle ͰιϩۜΛ֫ಘ͍ͯͨͨ͠Ίे෼ͳʮσʔλ෼ੳʯͷྗ͕͋Δͱ൑ அͯ͠ʮσʔλ෼ੳʯνʔϜͷΠϯλʔϯʹݺΜֶͩੜ͕͍·͍ͪ׆༂Ͱ͖ ͳ͔ͬͨɻ ͍ۤܦݧͨͪʢϑΟΫγϣϯʣ
6. ©2020 Wantedly, Inc. ʮσʔλ෼ੳʯͱ͍͏Ϟϊͷද͢ҙຯ͕ᐆດͳͨΊɼਓʹΑͬ ͯʮσʔλ෼ੳʯʹٻΊΔظ଴஋ʹζϨ͕ੜ͡Δɻ ໰୊
7. ©2020 Wantedly, Inc. ͋Δ෺ࣄΛ෼ղͯ͠ɺͦΕΒΛ੒ཱ͍ͤͯ͞Δ੒෼ɾཁૉɾଆ ໘Λ໌Β͔ʹ͢Δ͜ͱɻ ʮσʔλ෼ੳʯͱ͸ʁ: ʰ޿ࣙԓʱʹΑΔͱ… ʮ෼ੳʯ
8. ©2020 Wantedly, Inc. ʮσʔλ෼ੳʯͱ͸ʁ: Α͘ݟ͔͚Δ΍ͭ σʔλαΠΤϯςΟετʹٻΊΒΕΔεΩϧηοτ http://www.datascientist.or.jp/files/news/2014-12-10.pdf
9. ©2020 Wantedly, Inc. ʮσʔλ෼ੳʯͱ͸ʁ: ݕࡧͯ͠Έͨ όζϫʔυΛײ͡Δ
10. ©2020 Wantedly, Inc. ઈରతͳਖ਼ղͳΜͯʢ͖ͬͱʣͳ͍ͷͰࣗ෼ͳΓʹཧղɾղ ऍͯ͠ղ૾౓Λ্͛Δ͜ͱ͕ॏཁ͔ͳͱࢥͬͨɻ ͷͰɼ೥຤೥࢝ʹʮσʔλ෼ੳʯʹର͍ͯ͠Ζ͍Ζͱࢥ͍Λ஘ ͍ͤͯ·ͨ͠ɻ ʮσʔλ෼ੳʯͱ͸ʁ: ਖ਼ղͳΜͯͳ͍
11. ©2020 Wantedly, Inc. @yu-ya4 ͕Ұ൪εοΩϦͨ͠ʮσʔλ෼ੳʯͷ঺հ ※Ҿ༻͍ͯ͠·͕ࣗ͢෼ͳΓͷղऍ΋ೖΕͯ·͢
12. ©2020 Wantedly, Inc. ʰA or Bʱͱ͍͏໰͍ʹରͯ͠ղΛग़͢͜ͱ ʮσʔλ෼ੳʯͱ͸ʁ: Hikaru Kashida ͞Μ (@hik0107) ͷ note https://note.com/hik0107/n/ne240d47572c9?magazine_key=m3bf0289df2d6 ʮ෼ੳʯͱ͸ͻͱ͜ͱͰ͍͏ͱ…
13. ©2020 Wantedly, Inc. ʮYes or Noʯ ʰA or Bʱͱ͍͏໰͍ʹରͯ͠ղΛग़͢͜ͱ ʰA or Bʱͱ͍͏໰͍ ղͷܗࣜ ʮ͜ͷ༧ଌϞσϧͰ࡞ΒΕΔϥϯΩϯά͸ࠓΑΓྑ͍ʁʯ ʮ͜ͷUIมߋࢪࡦ͸ྑ͔ͬͨʁʯ ʮimpression or click or …ʯ ʮϢʔβͷܧଓ཰ͱ΋ͬͱ΋૬ؔͷߴ͍ߦಈ͸ʁʯ ʮYes or Noʯ
14. ©2020 Wantedly, Inc. ʮ͜Ε·ͰͷࢪࡦͰ͸͍͍ͤͥ105%ʹ͔͠ͳΒͳ͔ͬͨʯ ʮࠓ݄ͷ໨ඪ͸ PV ਺Λ140%ʹ͢Δ͜ͱʯ ʮࠓ݄தʹࢪࡦ͸͋ͱ3࣮ͭࢪͰ͖Δʯ ʰA or Bʱͱ͍͏໰͍ʹରͯ͠ղΛग़͢ͷʹඞཁͳཁૉ ࣄ࣮ = ਺ࣈ ൑அج४ ʮPV ਺͕ 120% ʹͳͬͨʯ ʰA or Bʱͱ͍͏໰͍ ʮ͜ͷUIมߋࢪࡦ͸ྑ͔ͬͨʁʯ ղ ʮYesʯ
15. ©2020 Wantedly, Inc. ʮࠓظͷ໨ඪ͸Ϣʔβͷܧଓ཰Λ্͛Δ͜ͱʯ ʮϢʔβͷܧଓ཰ͱ click ͸ڧ͍ਖ਼ͷ૬ؔΛ࣋ͭʯ ʮաڈʹΦϑϥΠϯධՁʹ͓͍ͯ MAP ͕105%ʹͳͬͨϥ ϯΩϯάʹΑͬͯclick ਺͕110%ʹͳͬͨ͜ͱ͕͋Δʯ ʰA or Bʱͱ͍͏໰͍ʹରͯ͠ղΛग़͢ͷʹඞཁͳཁૉ ࣄ࣮ = ਺ࣈ ൑அج४ ʮΦϑϥΠϯධՁʹ͓͍ͯ click Λਖ਼ղͱͨ͠ MAP ͕ 110% ʹͳͬͨʯ ʰA or Bʱͱ͍͏໰͍ ʮ͜ͷ༧ଌϞσϧͰ࡞ΒΕΔϥϯΩϯά͸ ࠓΑΓྑ͍ʁʯ ղ ʮYesʯ
16. ©2020 Wantedly, Inc. 1. ʰA or Bʱͱ͍͏໰͍ 2. ࣄ࣮ = ਺ࣈ 3. ൑அج४ 4. ղ ·ͱΊ:ʮσʔλ෼ੳʯͷཁૉ
17. ©2020 Wantedly, Inc. 1.ʰA or Bʱͱ͍͏໰͍ • ͜Ε͕ͳ͍ͱͦ΋ͦ΋ʮσʔλ෼ੳʯͰ͖ͳ͍ɻ • Ϗδωε՝୊Λ෼ੳ՝୊ʹམͱ͠ࠐΉ຋༁ऀͱ͔ݴΘΕΔ΍ͭʁϏδωεྗɻ • ந৅తͳʢࡶͳʣ໰͍Λ͜͜·ͰϒϨΠΫμ΢ϯͯ͠ઃܭ͢Δඞཁ͕͋Δ͜ͱ΋ɻ ຊ౰ʹ஌Γ͍ͨ͜ͱΛਪ͠ྔΔೳྗ͕ඞཁɻ • ex.ʮઌ݄ͷച্͸͍͘Βʁʯ-> ʮച্͸৳ͼ͍ͯΔͱݴ͑Δͷ͔ʁʯ • ΋ͪΖΜɼͦ΋ͦ΋σʔλΛूܭͯ͠ʮࣄ࣮ʹ਺ࣈʯΛग़͚ͩ͢Ͱ͍͍৔߹΋ɻ • ex. ϏδωενʔϜ͔ΒͷSQL࡞੒ґཔʢͷҰ෦ʣ
18. ©2020 Wantedly, Inc. 2. ࣄ࣮ = ਺ࣈ • ଘࡏ͢Δ࣮σʔλʹجͮ͘٬؍తࣄ࣮ɻҰ൪෼ੳͬΆ͍(?)ɻ • ͍ΘΏΔखΛಈ͔͢ೳྗ͕ඞཁɻ • SQL Ͱσʔλͷूܭ͢ΔೳྗɼKaggle (modeling)ྗɼetc. • σʔλΤϯδχΞϦϯάྗͱ͔΋͜͜ʹೖͬͯ͘Δ͔΋ʁਖ਼͘͠ʮࣄ࣮ = ਺ࣈʯΛ ಋग़Ͱ͖Δ؀ڥΛద੾ʹ࡞ΔͬͯҙຯͰɻ
19. ©2020 Wantedly, Inc. 3. ൑அج४ • ීวతͳ΋ͷͰ͸ͳ͘ɼͦͷ৔ʹ͓͚Δ͍͔ͭ͘ͷલఏ৚݅ʹج͖ͮಋग़͞ΕΔɻ • ͚ͬ͜͏͜͜Λ๨Ε͕ͪɻ • υϝΠϯ஌ࣝ௒ॏཁ • ࿦ཧྗ • ΈΜͳͷೲಘײ΋େ੾ɻ • ͜͜Λͳ͍͕͠Ζʹ͢ΔͱɼޡͬͨҙࢥܾఆΛͯ͠͠·͏ɻ΍͹͍ɻ
20. ©2020 Wantedly, Inc. 4. ղ • ʮʰA or Bʱͱ͍͏໰͍ʯ΁ͷճ౴ • ʮ൑அج४ʯʹج͖ͮʮࣄ࣮ = ਺ࣈʯ͔ΒʮʰA or Bʱͱ͍͏໰͍ʯʹରͯ͠ಋग़ ͢Δɻ • ՄࢹԽͳͲΛۦ࢖ͯ͠આ໌Մೳͳܗʹ͢Δɻਓʹ఻͑Δೳྗ͕ඞཁɻ • ผͷԾઆɼ࣍ͷʮʰA or Bʱͱ͍͏໰͍ʯͷώϯτʢʹʮσʔλ෼ੳʯͷλωʣΛ ݟग़͢ೳྗ͕͋Δͱڧ͍ɻ
21. ©2020 Wantedly, Inc. ͜ΕΒ4ͭͷཁૉΛʮσʔλ෼ੳʯʹؔ͢Δීஈͷۀ຿΍νʔϜฤ੒ɼ ҭ੒ɼ࠾༻ͳͲͷࡍʹҙࣝͯ͠ɼͲͷཁૉʹ஫໨͢΂͖ͳͷ͔ߟ͑Ε͹ ʮσʔλ෼ੳʯ΁ͷظ଴஋ͷζϨΛݮΒͤΔ͔΋ʁ ·ͱΊ:ʮσʔλ෼ੳʯͷཁૉ 1. ʰA or Bʱͱ͍͏໰͍ 2. ࣄ࣮ = ਺ࣈ 3. ൑அج४ 4. ղ
22. ©2020 Wantedly, Inc. ͜ΕΒ4ͭͷཁૉΛʮσʔλ෼ੳʯʹؔ͢Δීஈͷۀ຿΍νʔϜฤ੒ɼ ҭ੒ɼ࠾༻ͳͲͷࡍʹҙࣝͯ͠ɼͲͷཁૉʹ஫໨͢΂͖ͳͷ͔ߟ͑Ε͹ ʮσʔλ෼ੳʯ΁ͷظ଴஋ͷζϨΛݮΒͤΔ͔΋ʁ ·ͱΊ:ʮσʔλ෼ੳʯͷཁૉ 1. ʰA or Bʱͱ͍͏໰͍ 2. ࣄ࣮ = ਺ࣈ 3. ൑அج४ 4. ղ ※ࢲʹͱͬͯ͸͜ͷղऍ͕ͬ͘͠Γ͖͚ͨͩͰɼਖ਼ղ͸ͳ͍ɻ
23. ©2020 Wantedly, Inc. ʮσʔλ෼ੳʯͱ͍͏ந৅తͳϞϊͷղ૾౓Λ্͛ͯʢࣗ෼ ͳΓͷղऍͰ͍͍ͷͰʣ໌֬ͳఆٛΛ࣋ͯΕ͹ɼ΋͏গ͠ ʮσʔλ෼ੳʯͱ͏·͖ͭ͋͑͘Δ͔΋͠Ε·ͤΜɻ ࠓ೔఻͔͑ͨͬͨ͜ͱ
24. ©2020 Wantedly, Inc. ͞Βʹɼࣾ಺ͳ͍͠͸νʔϜʹ͓͍ͯʮσʔλ෼ੳʯͱ͍͏Ϟϊʹ ରͯ͠ڞ௨ͷೝࣝΛ࣋ͯ͹΋ͬͱ޾ͤʹͳΕΔ͔΋͠Ε·ͤΜɻ ࠓ೔఻͔͑ͨͬͨ͜ͱ(+α)
25. ©2020 Wantedly, Inc. We are hiring!! https://www.wantedly.com/projects/375150