会社訪問アプリ「Wantedly Visit」のデータで見る相互推薦システム / deim2021-rrs-wantedly-visit

Slide 1

Slide 1 text

Slide 2

Slide 2 text

©2021 Wantedly, Inc. 1. ͸͡Ίʹ • ࣗݾ঺հɺձࣾͱϓϩμΫτͷ঺հ • ϓϩμΫτʹ͓͚ΔσʔλαΠΤϯεͷऔΓ૊Έࣄྫ • ΞΧσϛΞʹ͓͚Δ׆ಈ 2. ૬ޓਪનγεςϜͱ͸ • ૬ޓਪનγεςϜͷ֓ཁɾಛ௃ • طଘख๏ͷ঺հ 3. ձࣾ๚໰ΞϓϦʮWantedly VisitʯͷσʔλͰݟΔ૬ޓਪનγεςϜ • Wantedly Visit ͷσʔλʹରͯ͠طଘख๏Λద༻࣮ͨ͠ݧɾߟ࡯ • ࣮ݧ݁ՌΛड͚ͨվྑख๏ͷఏҊɾ࣮ݧɾߟ࡯ ձࣾ๚໰ΞϓϦʮWantedly VisitʯͷσʔλͰݟΔ૬ޓਪનγεςϜ

Slide 3

Slide 3 text

Slide 4

Slide 4 text

Slide 5

Slide 5 text

Slide 6

Slide 6 text

Slide 7

Slide 7 text

Slide 8

Slide 8 text

Slide 9

Slide 9 text

Slide 10

Slide 10 text

Slide 11

Slide 11 text

Slide 12

Slide 12 text

Slide 13

Slide 13 text

Slide 14

Slide 14 text

Slide 15

Slide 15 text

©2021 Wantedly, Inc. User Item ैདྷͷҰൠతͳਪનγεςϜ ૬ޓਪનγεςϜ User(Female) User(Male) User(Job Seeker) User(Recruiter/Company) ex. Amazon, Netflix ex. Tinder, Pairs ex. Wantedly, LinkedIn ૬ޓਪનγεςϜʢReciprocal Recommender Systemsʣͱ͸ʁ ʮαʔϏε಺ͷϢʔβΛޓ͍ʹਪન͢ΔγεςϜʯ ਪન ਪન ਪન

Slide 16

Slide 16 text

Slide 17

Slide 17 text

©2021 Wantedly, Inc. User A to User B Preference Score User B to User A Preference Score Reciprocal Preference Score Aggregation 1. γεςϜ಺ͷϢʔβ͸ A ͱ B ͷ̎ͭͷάϧʔϓʹ෼͔Ε͓ͯΓɺҟͳΔάϧʔϓͷϢʔβ͕ޓ͍ਪન͞ΕΔ ΋ͷͱ͢Δɻʢe.g. σʔςΟϯάαʔϏεʹ͓͚ΔஉঁɺٻਓαʔϏεʹ͓͚Δٻ৬ऀͱاۀʣ 2. ୯ํ޲ͷᅂ޷ͷେ͖͞Λද͢ Preference Score ΛɺA ͔Β B ΁ͷϢʔβٴͼ B ͔Β A ͷϢʔβͷͦΕͧΕ ʹ͍ͭͯܭࢉ 3. Aggregation Function Λར༻ͯ͠ɺ̎ͭͷ Preference Score Λ૊Έ߹Θͤͯ૒ํ޲ͷᅂ޷ͷେ͖͞Λද͢ Reciprocal Preference Score Λܭࢉ ૬ޓਪનγεςϜʹ͓͚Δᅂ޷ͷ༧ଌ

Slide 18

Slide 18 text

©2021 Wantedly, Inc. ✓ [RECON] (Pizzato 2010) ‣ ϢʔβͷϓϩϑΟʔϧ৘ใΛར༻ͨ͠ίϯςϯπϕʔεϑΟϧλϦϯάͰ Preference Score Λࢉग़ ‣ Aggregation Function ʹ͸ௐ࿨ฏۉΛར༻ ✓ [RCF] (Xia 2015) ‣ ߦಈཤྺʹجͮ͘ϝϞϦϕʔεͷϢʔβϕʔεڠௐϑΟϧλϦϯάʢk-ۙ๣ʣͰ Preference Score Λࢉग़ ✓ [LFRR](Neve 2019) ‣ ߦಈཤྺʹج͖ͮ࡞੒ͨ͠ User-User ߦྻʹ Matrix Factorization Λద༻ͯ͠ Preference Score Λࢉग़ ‣ Aggregation Function ʹ͍ͭͯɼௐ࿨ฏۉҎ֎ͷؔ਺ʹ͍ͭͯ΋ൺֱ࣮ݧ ✓ [ImRec](Neve 2020) ‣ ϢʔβͷϓϩϑΟʔϧը૾Λར༻ͨ͠ίϯςϯπϕʔεϑΟϧλϦϯάͰ Preference Score Λࢉग़ ૬ޓਪનγεςϜʹ͓͚Δᅂ޷ͷ༧ଌͷطଘख๏

Slide 19

Slide 19 text

Slide 20

Slide 20 text

Slide 21

Slide 21 text

©2021 Wantedly, Inc. • طଘख๏ʹ͓͚ΔධՁ࣮ݧ͸σʔςΟϯάαʔϏεʹ͓͚Δ΋ͷ͕ଟ͍ɻ ➡ ٻਓαʔϏεͷ࿮૊ΈͰ͋Δ Wantedly Visit ͷσʔλͰͷ࣮ݧʹҰఆͷՁ஋͕͋ΔͷͰ͸ʁ • Wantedly Visit ͸ҰൠతͳٻਓαʔϏεΑΓ΋ΧδϡΞϧʹϢʔβ͕ߦಈ͢Δ αʔϏεͰ͋ΔͨΊɺൺֱతϢʔβͱاۀؒͷߦಈϩά͕ଟ͍ɻ ➡ ڠௐϑΟϧλϦϯάϕʔεͷطଘख๏Λͦͷ··ద༻ͯ͠΋Ұఆͷੑೳ͕ग़ΔͷͰ͸ʁ • ݱࡏͷطଘख๏͸ൺֱతφΠʔϒͳ΋ͷͰ͋ΔͨΊɺվྑͷ༨஍͕͋Δɻ ➡ طଘख๏Ͱͷ࣮ݧΛ௨ͯ͠վྑख๏ΛఏҊͰ͖ΔͷͰ͸ʁ Ϟνϕʔγϣϯ

Slide 22

Slide 22 text

Slide 23

Slide 23 text

©2021 Wantedly, Inc. • Preference Score ༧ଌܭࢉͷͨΊͷΞϧΰϦζϜ • ߦಈཤྺʹجͮ͘ϝϞϦϕʔεͷϢʔβϕʔεڠௐϑΟϧλϦϯά [RCF] (Xia 2015) • ߦಈཤྺʹج͖ͮ࡞੒ͨ͠ User-User ߦྻʹ Matrix Factorization Λద༻ [LFRR](Neve 2019) • Aggregation Function • (Neve 2019)Ͱ࣮ݧ͞Ε͍ͯͨ4छྨ • Arithmetic Mean (AM) • Geometric Mean (GM) • Harmonic Mean (HM) • Cross-Ratio Uninorm (CRU) ࣮ݧ͢Δطଘख๏ User A to User B Preference Score User B to User A Preference Score Reciprocal Preference Score Aggregation Function CRU:

Slide 24

Slide 24 text

©2021 Wantedly, Inc. ࣮ݧ֓ཁ Wantedly Visit ʹ͓͚ΔϢʔβͱاۀͷ Matching Λ༧ଌ Company User Ԡื ϝοηʔδฦ৴ Company User εΧ΢τૹ৴ ϝοηʔδฦ৴ Company User Ԡื ӾཡͷΈ Company User εΧ΢τૹ৴ ӾཡͷΈ • Ϣʔβ͕Ԡื or اۀ͕εΧ΢τૹ৴ͨ͠ࡍʹ૬ख͕ϝοηʔδΛฦ৴͢Ε͹ Match (positive) • Ԡื or εΧ΢τૹ৴Λ૬ख͕Ӿཡ্ͨ͠ͰϝοηʔδΛฦ৴͠ͳ͚Ε͹ Not Match (negative) Match Not Match

Slide 25

Slide 25 text

©2021 Wantedly, Inc. ᅂ޷σʔλ/ධՁ஋ • Ϣʔβͷᅂ޷σʔλ • اۀ΁ͷԠื • اۀ͔ΒͷεΧ΢τૹ৴ʹର͢Δϝοηʔδฦ৴ • اۀͷᅂ޷σʔλ • Ϣʔβ΁ͷεΧ΢τૹ৴ • Ϣʔβ͔ΒͷԠืʹର͢Δϝοηʔδฦ৴ • ධՁ஋: Ϣʔβͱاۀͷ૊ʹରͯ͠༩͑ΒΕΔ (boolean) • Ϣʔβ͕Ԡื or اۀ͕εΧ΢τૹ৴ͨ͠ࡍʹ૬ख͕ϝοηʔδΛฦ৴͢Ε͹ Match (positive) • Ԡื or εΧ΢τૹ৴Λ૬ख͕Ӿཡ্ͨ͠ͰϝοηʔδΛฦ৴͠ͳ͚Ε͹ Not Match (negative) • Ԡื or εΧ΢τૹ৴͕ൃੜ͕ͨ͠૬ख͕Ӿཡ͍ͯ͠ͳ͍΋ͷͷධՁ஋͸ෆ໌ʢະධՁʣ

Slide 26

Slide 26 text

Slide 27

Slide 27 text

Slide 28

Slide 28 text

©2021 Wantedly, Inc. ࣮ݧ݁Ռ AUC AM GM HM CRU RCF 0.558 0.618 0.639 0.616 LFRR 0.475 0.578 0.622 0.552 RCF LFRR • طଘݚڀͱಉ༷ɺHM(ௐ࿨ฏۉ)͕΋ͬͱ΋ߴ͍ੑೳʹ • طଘݚڀͱҟͳΓɺLFRR ΑΓ΋ RCF ͷํ͕ߴ͍ੑೳʹ • طଘख๏ͷ࿦จͷ࣮ݧͱൺ΂ͯ΋ܦݧతʹ΋ AUC ͷ஋͕ খ͍͞ Aggregation Function Algorithm

Slide 29

Slide 29 text

Slide 30

Slide 30 text

©2021 Wantedly, Inc. ߟ࡯ɾԾઆ طଘख๏ͷ࿦จͷ࣮ݧͱൺ΂ͯ΋ܦݧతʹ΋ AUC ͷ஋͕খ͍͞ • Ϣʔβ͔Βاۀɺاۀ͔ΒϢʔβͱ͍͏ҟͳΔ Preference Score Λ୯७ͳ Aggregation Function ʹ͔͚͍ͯΔͷ͕ྑ͘ͳ͍ʁ • ͦΕͧΕͷ෼෍΍εέʔϧ͕ҟͳΔ • ಉ͡஋Λऔ͍ͬͯΔ͔Βͱ͍ͬͯɺಉ͘͡Β͍ͷᅂ޷ͷେ͖͞Λද͢ͷ͔ʁ • ୯ํ޲ͷ Preference Score ͷ༧ଌͷ࣌఺Ͱਫ਼౓͕ྑ͘ͳ͍ʁ

Slide 31

Slide 31 text

Slide 32

Slide 32 text

©2021 Wantedly, Inc. ߟ࡯ɾԾઆ ୯ํ޲ͷ Preference Score ͷ༧ଌͷ࣌఺Ͱਫ਼౓͕ྑ͘ͳ͍ʁ • Ϣʔβͷ༧ଌ Preferenc Score ٴͼ اۀͷ༧ଌ Preference Score ΛͦΕͧΕධՁ AUC Ϣʔβ اۀ RCF 0.850 0.737 LFRR 0.833 0.727 Subject Algorithm → اۀͷ Preference Score ͷ༧ଌਫ਼౓͕௿͍

Slide 33

Slide 33 text

©2021 Wantedly, Inc. ߟ࡯ɾԾઆ ͳͥاۀͷ Preference Score ͷ༧ଌਫ਼౓͕௿͍ͷ͔ • اۀͷᅂ޷σʔλ͸Ϣʔβʹൺ΂ͯগͳ͍ • اۀͷᅂ޷Λڧ͘ද͢ೳಈతͳᅂ޷σʔλʢεΧ΢τૹ৴ʣͷର৅ͱͳΔϢʔβ਺͕গͳ͍ • डಈతͳᅂ޷σʔλΑΓ΋ೳಈతͳᅂ޷σʔλͷํ͕ڧ͘Ϣʔβͷᅂ޷Λද͢ͷͰ͸ͳ͍͔ʁ • اۀͷडಈతͳᅂ޷σʔλʢϢʔβͷԠืʹର͢Δϝοηʔδฦ৴ʣͷର৅ͱͳΔϢʔβ਺ ͸ଟ͍͕ɺ͋·Γᅂ޷Λද͍ͤͯͳ͍ • Ϣʔβʹൺ΂ͯɺاۀ͸͋·Γᅂ޷ʹ߹͍ͬͯͳ͍Ϣʔβʹରͯ͠΋ϝοηʔδฦ৴Λߦ͏ʁ

Slide 34

Slide 34 text

Slide 35

Slide 35 text

Slide 36

Slide 36 text

Slide 37

Slide 37 text

©2021 Wantedly, Inc. ఏҊख๏ᶄ ෛͷᅂ޷σʔλΛ׆༻͢ΔʢNegʣ • Ϣʔβ/اۀͷᅂ޷ΛΑΓਖ਼֬ʹ༧ଌ͢ΔͨΊɺԠื/εΧ΢τૹ৴ ͞ΕͯӾཡ͕ͨ͠ϝο ηʔδฦ৴͠ͳ͔ͬͨͱ͍͏ෛͷᅂ޷σʔλΛར༻ • ͜Ε·Ͱͷਖ਼ͷᅂ޷σʔλΛར༻ͨ͠ਖ਼ͷ Preference Score ͷ༧ଌʹՃ͑ɺෛͷ Preference Score Λܭࢉ͢Δɻ • ਖ਼ͱෛͷ༧ଌ Preference Score Λ଍͠߹Θͤͯ༧ଌ Preference Score Λܭࢉ

Slide 38

Slide 38 text

©2021 Wantedly, Inc. ࣮ݧ݁Ռɾߟ࡯ʢ୯ํ޲ͷᅂ޷: Preference Scoreʣ • اۀͷ Preference Score ͷ༧ଌਫ਼౓͕େ͖͘վળ • Ϣʔβͷ Preference Score ͷ༧ଌਫ਼౓͸اۀʹൺ΂Δͱ͋·Γվળͤͣ • Ϣʔβ͸ᅂ޷ʹؔ܎ͳ͘ϝοηʔδΛฦ৴͠ͳ͍͜ͱ͕ଟ͍ͨΊɺෛͷ Preference Score ͷޮՌ ͕খ͍͞ʁʢe.g. ΊΜͲ͍͔͘͞Βϝοηʔδฦ৴͠ͳ͍ʣ AUC Ϣʔβ اۀ RCF 0.850 0.737 LFRR 0.833 0.727 LFRR + Scaler + Neg 0.850 0.817 Subject Algorithm

Slide 39

Slide 39 text

©2021 Wantedly, Inc. ࣮ݧ݁Ռɾߟ࡯ʢMatching: Reciprocal Preference Scoreʣ AUC AM GM HM CRU RCF 0.558 0.618 0.639 0.616 LFRR 0.475 0.578 0.622 0.552 LFRR + Scaler 0.536 0.619 0.639 0.603 LFRR + Scaler + Neg 0.543 0.686 0.712 0.674 LFRR + Scaler + Neg Aggregation Function Algorithm • ఏҊख๏Ͱ͋Δ LFRR + Scaler + Neg ͱ HM ͷ૊Έ߹Θ͕ͤ࠷΋ߴ͍ੑೳʹ • Scaler ͷΈͷద༻Ͱ΋༧ଌੑೳͷ޲্͕֬ೝͰ͖Δ • ґવɺAUC ͸͞΄Ͳେ͖͘ͳ͍ɻ • Preference Score ͷ༧ଌܭࢉ΍ Aggregation Function ʹ΋ͬͱߴ౓ͳΞϧΰϦζϜΛར༻͢Δɻ • ೳಈతͳᅂ޷σʔλͱडಈతͳᅂ޷σʔλͷॏΈΛมֶ͑ͯश͢Δɻ

Slide 40

Slide 40 text

©2021 Wantedly, Inc. ·ͱΊ A. Wantedly Visit ͷσʔλʹରͯ͠طଘख๏Λద༻࣮ͨ͠ݧɾߟ࡯ • ઌߦݚڀͱྨࣅ͢ΔΑ͏ͳ࣮ݧ݁Ռ • શମతʹ༧ଌਫ਼౓͕௿͍ • Ϣʔβ͔Βاۀɺاۀ͔ΒϢʔβͱ͍͏ҟͳΔ Preference Score Λ୯७ͳ Aggregation Function ʹ ͔͚͍ͯΔ͜ͱ͕ྑ͘ͳ͍ʁ • اۀͷ Preference Score ͷ༧ଌਫ਼౓͕௿͍͜ͱ͕ྑ͘ͳ͍ʁ B. ࣮ݧ݁ՌΛड͚ͨվྑख๏ͷఏҊɾ࣮ݧɾߟ࡯ • ख๏ͷఏҊ • Ϣʔβͱاۀͷ Preference Score ͷεέʔϧΛ߹ΘͤΔʢScalerʣ • ෛͷᅂ޷σʔλΛ׆༻͢ΔʢNegʣ • ఏҊख๏͕طଘख๏ͷੑೳΛ্ճͬͨ

Slide 41

Slide 41 text

©2021 Wantedly, Inc. 3FGT • (Pizzato 2010) Luiz Pizzato, Tomek Rej, Thomas Chung, Irena Koprinska, and Judy Kay. 2010. RECON: a reciprocal recommender for online dating. Proceedings of the fourth ACM conference on Recommender systems P. 207-214. • (Pizzato 2012) Luiz Pizzato, Tomasz Rej, Joshua Akehurst, Irena Koprinska, Kalina Yacef, and Judy Kay. 2012. Recommending people to people: the nature of reciprocal recommenders with a case study in online dating. User Model User-Adap Inter (2013) 23: 447. • (Xia 2015) Peng Xia, Benyuan Liu, Yizhou Sun, and Cindy Chen. 2015. Reciprocal Reciprocal recommendation System for Online Dating. Proceedings of the 2015 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining P. 234-241. • (Neve 2019) J Neve, I Palomares.2019. Latent Factor Models and Aggregation Operators for Collaborative Filtering in Reciprocal Recommender Systems Proceedings of the 13th ACM Conference on Recommender Systems, 219-227. • (Neve 2020) J Neve, R McConville.2020. ImRec: Learning Reciprocal Preferences Using Images. Proceedings of the 14th ACM Conference on Recommender Systems, 170-179.