Upgrade to Pro — share decks privately, control downloads, hide ads and more …

Review: "Recommending Investors for Crowdfundin...

yag_ays
July 09, 2014

Review: "Recommending Investors for Crowdfunding Projects"

yag_ays

July 09, 2014
Tweet

More Decks by yag_ays

Other Decks in Research

Transcript

  1. ࠓճ঺հ͢Δ࿦จͷ֓ཁ • “Recommending Investors for Crowdfunding Project” [Jisun+ 2014] •

    WWW 2014 (Seoul, KOREA) • Jisun AnͷYahoo Labs in Barcelona Πϯλʔϯγοϓͷ࢓ࣄ ! • ࠷ऴతͳ໨ඪɿKickstarterͷϑΝ΢ϯμʔͱग़ࢿऀͷϚονϯά • KickstarterͷϓϩδΣΫτ΍ग़ࢿऀͷ෼ੳϝΠϯͳͱ͜Ζ͕͋Δ 2
  2. • Ϋϥ΢υϑΝϯσΟϯά • 2012೥ʹूΊͨ౤ࢿ૯ֹ$320 million • c.f. ౤ࢿՈ/ϕϯνϟʔΩϟϐλϧʹΑΔग़ࢿ ! •

    ϑΝ΢ϯμʔ͸໨ඪֹۚΛઃఆͯ͠౤ࢿΛืΔ • ౤ࢿՈ͸౤ࢿֹۚʹԠͯ͡ใु͕΋Β͑Δ • e.g. $100౤ࢿͰ੡඼1ݸϓϨθϯτɼ$300౤ࢿͰ੡඼5ݸϓϨθϯτ Kickstarterͱ͸ https://www.kickstarter.com/help/style_guide 3
  3. Kickstarter੒ޭࣄྫ • Oculus Rift • 9,522 / $ 2,437,429 •

    Memoto (Narrative Clip) • 2,871 / $ 550,189 • Little Witch Academia 2 • 7,938 / $ 625,518 https://www.kickstarter.com/projects/1523379957/oculus-rift-step-into-the-game https://www.kickstarter.com/projects/martinkallstrom/memoto-lifelogging-camera https://www.kickstarter.com/projects/1311401276/little-witch-academia-2 4
  4. Kickstarterͷಛघੑɿ౤ࢿʹࣦഊͨ͘͠ͳ͍͚Ͳ… • All or Nothing • ໨ඪֹۚʹୡ͠ͳ͚Ε͹ϓϩδΣΫτ͸ࣦഊɼ౤ࢿۚ͸શֹฦ٫ • ϓϩδΣΫτͷ੒ޭ/ࣦഊʹؔΘΒͣɼ౤ࢿऀ͸ଛΛ͠ͳ͍ !

    • ιʔγϟϧͳଆ໘ • ॳظ౤ࢿʹ͸਌଒΍༑ୡ͕ଟ͍ʢ20-40%ͱ͍͏ࢉग़΋ʣ • े෼ͳ౤ࢿՈΛूΊΒΕͳ͍ͱϓϩδΣΫτ͕ࣦഊ͠΍͍͢ 5
  5. σʔληοτ • Kickstarter͔ΒΫϩʔϧ • 2013೥7݄͔Β10݄ʹొ࿥͞Εͨ΋ͷ • USA಺෦ͷϑΝϯυͷΈ • ߹ܭ 1,149ϓϩδΣΫτ/

    78,460ग़ࢿऀ • Twitter͔ΒΫϩʔϧ • ϓϩδΣΫτʹݴٴ͢ΔtweetͷΈ • ߹ܭ71,315 tweetΛऩू 8 (Average)
  6. ౤ࢿՈͷߏ੒ • ߹ܭ78,460ਓͷग़ࢿऀ ! • 4ճະຬͷ౤ࢿΛͨ͠ਓ͸51% • ؾ·͙Ε౤ࢿऀ “Occasional Investors”

    • 32ճҎ্ͷ౤ࢿΛͨ͠ਓ͸11% • ৗ࿈౤ࢿऀ “Frequent Investors” ؾ·͙Ε౤ࢿՈ ৗ࿈౤ࢿՈ 9
  7. ౤ࢿՈͷڍಈʹؔ͢ΔԾઆ • ৗ࿈౤ࢿՈ͸ҎԼͷΑ͏ͳੑ࣭ͷϓϩδΣΫτʹ౤ࢿ͠΍͍͢ • ৘ใ͕සൟʹΞοϓσʔτ͞ΕΔ • ϑΝ΢ϯμʔ͕౤ࢿՈͷ࣭໰ʹ౴͑Δ • ౤ࢿͷใु͕ྑ͍ •

    ߴ͍౤ࢿֹۚͷϓϩδΣΫτ͸ৗ࿈౤ࢿՈʹ౤ࢿ͞Ε΍͍͢ • ϩʔΧϧͳϓϩδΣΫτ͸ؾ·͙Ε౤ࢿՈʹ౤ࢿ͞Ε΍͍͢ • ૣ͘౤ࢿΛूΊΔϓϩδΣΫτ͸ৗ࿈౤ࢿՈʹ౤ࢿ͞Ε΍͍͢ • ৗ࿈౤ࢿՈ͸ࣗ෼ͷڵຯ͋ΔϓϩδΣΫτʹ౤ࢿ͠΍͍͢ 11
  8. Ծઆɿ౤ࢿՈͷڵຯͱ౤ࢿઌͷؔ܎ • LDA (Latent Dirichlet Allocation) Λ࢖ͬͨτϐοΫͷྨࣅ౓ • ౤ࢿͨ͠ϓϩδΣΫτͷ֓ཁͱ౤ࢿՈͷTweetͷ಺༰ (200

    tweetsఔ౓) • (τϐοΫ਺͸Ṗ) ! • ৗ࿈౤ࢿՈ͸ࣗ෼ͷڵຯ͋Δෳ਺ͷτϐοΫʹ౤ࢿ͕ͪ͠ • ؾ·͙Ε౤ࢿՈ͸ࣗ෼ͷڵຯͱ͸ؔ܎ͳ͍τϐοΫʹ౤ࢿ or ͻͱͭͷτ ϐοΫʹภͬͯ౤ࢿ͍ͯ͠Δ 15
  9. ͜͜·Ͱͷ·ͱΊ • ৗ࿈౤ࢿՈ (4ϲ݄ؒʹ32ճҎ্౤ࢿͨ͠Frequent Investors) • Α͘Ϛωʔδϝϯτ͞Εɼ໨ඪֹ͕ۚߴ͘ɼࣗ෼ͷڵຯʹ߹͏ϓϩδΣ Ϋτʹ౤ࢿ͢Δ܏޲ʹ͋Δ • ௨ৗͷ౤ࢿՈͷΑ͏ͳڍಈΛࣔ͢

    • ؾ·͙Ε౤ࢿՈ (4ϲ݄ؒʹ4ճະຬͷ౤ࢿΛͨ͠Occasional Investors) • ࠓճબΜͩಛ௃ྔʹ͋·ΓӨڹ͞Εͣ౤ࢿΛߦ͏ • ౤ࢿͱ͍͏ΑΓ΋د෇ͱ͍͏ײ͡ 16
  10. Ϩίϝϯσʔγϣϯͷख๏ͱධՁํ๏ • ੑೳධՁ͢Δख๏4ͭ : {LR, SVM-linear, SVM-poly, SVM-RBF} • ϩδεςΟοΫճؼʢLRʣ

    • 3छྨͷΧʔωϧΛ༻͍ͨSVM ʢLinear, polynomial, RBFʣ ! • ධՁɿ5-fold cross validation • σʔληοτͷ80%Ͱֶशˠ20%ͰධՁ Λ5ճ܁Γฦͯ͠ධՁΛฏۉ 20
  11. ࢖༻͢Δಛ௃ྔ • Static Feature: ϓϩδΣΫτൃ଍࣌ʹΘ͔Δಛ௃ྔ • ໨ඪֹۚɾใुͷϨϕϧɾաڈʹ౤ࢿͨ͠ϓϩδΣΫτͷΧςΰϦɾ Twitterͷ౤ߘʹΑΔ౤ࢿՈͷڵຯ • Dynamic

    Feature: ϓϩδΣΫτͷਐߦʹΑͬͯ൑໌ͯ͘͠Δಛ௃ྔ • ϓϩδΣΫτͷ੒ޭ཰ɾߋ৽਺ɾίϝϯτ਺ɾ஍ཧతͳڑ཭ͷ͹Β͖ͭ 21
  12. ϨίϝϯσʔγϣϯͷධՁ • RBFΧʔωϧΛ༻͍ͨSVM • Static͚ͩͷಛ௃ྔ ɿ82% • Dynamic͚ͩͷಛ௃ྔɿ73% ! •

    StaticͱDynamicΛ߹ΘͤΔͱACC 84% ! • ਖ਼ྫෛྫͷׂ߹͕50-50ɿϕʔεϥΠϯ͸50% ACC : Accuracy P : Precision R : Recall F1 : F-score AUC : ROCۂઢԼͷ໘ੵ 22
  13. Ͳͷಛ௃ྔ͕ޮ͍͍ͯΔͷ͔ʁ • C : ίϝϯτ਺ • R : ใुͷϨϕϧ •

    S : ஍ཧతͳڑ཭ͷ͹Β͖ͭ • G: ੒௕཰ • E : ΧςΰϦʔͷҰக౓ • TS: ڵຯ͋ΔτϐοΫͱͷྨࣅ౓ →EͱTS͕ਫ਼౓޲্ʹد༩͍ͯ͠Δ 23
  14. ࿦จͷ·ͱΊ • ౤ࢿՈʹΑͬͯKickstarterͷ౤ࢿελΠϧ͕ҧ͏ • ৗ࿈౤ࢿՈ͸໺৺తͳϓϩδΣΫτʹ౤ࢿ͢Δ • ؾ·͙Ε౤ࢿՈ͸د෇ײ֮Ͱ౤ࢿ / ܳज़ʹؔ࿈ͨ͠ϓϩδΣΫτʹ౤ࢿ •

    ౤ࢿՈͱϓϩδΣΫτͷϚονϯά͸Մೳ • ౤ࢿՈ͕ϓϩδΣΫτʹ౤ࢿΛ͢Δ͔Ͳ͏͔͸84%ͷਫ਼౓ͰਪଌՄೳ • ౤ࢿՈͷڵຯ͋ΔΧςΰϦʔ΍಺༰͕Ϛονϯάʹڧ͘Өڹ͢Δ 24