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

有価証券報告書のテキストデータを用いた株価予測モデルと投資モデルの構築

 有価証券報告書のテキストデータを用いた株価予測モデルと投資モデルの構築

同志社大学文化情報学部卒業論文諮問会で使用した発表資料です.

hideyoshikato

March 20, 2021
Tweet

More Decks by hideyoshikato

Other Decks in Research

Transcript

  1. 5 2 mB 3 B 5 b 1, 1 4

    le 5 2, , 3, mB B 2 mB g g 4, B g 5, / 4 3 o P3 18 P20 21 P41 42 P44 45 P23 39
  2. 5 2 mB 3 B 5 b 1, 1 4

    le 5 2, , 3, mB B 2 mB g g 4, B g 5, / 4 3 o P3 18 P20 21 P41 42 P44 45 P23 39
  3. . 4 . / . 1    

        PER PBR 6 5 /
  4. . 47 . / . 1 $# 1  

    2 2 PER ROE 5 7  + "  ,&%*' !#0-   )( ./ /
  5. . 4 8 . / 8 . 1 +*&; 

    < < PER ROE 5 RSI %5-4 + *( ,%5 / %5)  6/ .30# '*:7 ! $ 21 8"9
  6. 9 5 . 4 [1] , , , & .

    (2018). 5 1 .  , 33(1), A-H51_1. [2] , & . (2015). 5 / 4 . 77 , 2015(1), 417-418. ϑΝϯμϝϯλϧ෼ੳͱςΫχΧϧ෼ੳͷ྆ํΛߟྀͨ͠ಛ௃ྔΛ༻͍ͯגՁมಈ Λ༧ଌ͠ചങΛߦͬͨ݁Ռɼैདྷͷചങख๏ΑΓ΋ߴ͍ऩӹ཰ΛಘΒΕΔͱࣔͨ͠ <௕ඌɾ௕ඌ > ϑΝϯμϝϯλϧ෼ੳʹՃ͑ͯςΫχΧϧ෼ੳΛ૊Έ߹ΘͤͨఏҊϞσϧ͕֤୯ମ ͷख๏ΑΓߴ͍ਫ਼౓ͷגओՁ஋ਪఆΛ͓͜ͳ͑Δ͜ͱΛࣔͨ͠ <ాଜଞ >       .
  7. . . / 1 . 5 [1] , , ,

    & . (2018). 4 . .  , 33(1), A-H51_1. [2] , & . (2015). 0 5 . . 77 , 2015(1), 417-418. ϑΝϯμϝϯλϧ෼ੳͱςΫχΧϧ෼ੳͷ྆ํΛߟྀͨ͠ಛ௃ྔΛ༻͍ͯגՁมಈ Λ༧ଌ͠ചങΛߦͬͨ݁Ռɼैདྷͷചങख๏ΑΓ΋ߴ͍ऩӹ཰ΛಘΒΕΔɽ <௕ඌɾ௕ඌ > ϑΝϯμϝϯλϧ෼ੳʹՃ͑ͯςΫχΧϧ෼ੳΛ૊Έ߹ΘͤͨఏҊϞσϧ͕֤୯ମ ͷख๏ΑΓߴ͍ਫ਼౓ͷגओՁ஋ਪఆΛ͓͜ͳ͑Δɽ <ాଜଞ >       / 1
  8. 2 . 5 5 4 5/ 2   

     1 [ 2018] . 5
  9. 45 3 . ৽ฉهࣄͷ࣌ܥྻςΩετσʔλ͕גࣜࢢ৔ͷಈ޲Λ༧ଌ͢Δࡍʹ༗ޮͰ͋Δ͜ͱ Λࣔͨ͠ɽ <দҪɾ࿨ઘ > ೔ຊܦࡁ৽ฉͷ࣌ܥྻςΩετσʔλΛ༻͍ͯ501*9&5'ͷऔҾՁ֨ͷಅམΛ༧ଌ ͠ɼఏҊख๏͕༗༻Ͱ͋Δ͜ͱΛࣔͨ͠ɽ <দຊɾদҪ

    > [5] , & . (2018). 3 /1 1 . In   2018   ( 32 )  (pp. 2J201-2J201). .      [4] , & . (2016). 45 /1 . In 2016 ( 30 ) (pp. 3L3OS16a6-3L3OS16a6). . .
  10. 45 . ৽ฉهࣄͷ࣌ܥྻςΩετσʔλ͕גࣜࢢ৔ͷಈ޲Λ༧ଌ͢Δࡍʹ༗ޮͰ͋Δ͜ͱ Λࣔͨ͠ɽ <দҪɾ࿨ઘ > ೔ຊܦࡁ৽ฉͷ࣌ܥྻςΩετσʔλΛ༻͍ͯ501*9&5'ͷऔҾՁ֨ͷಅམΛ༧ଌ ͠ɼఏҊख๏͕༗༻Ͱ͋Δ͜ͱΛࣔͨ͠ɽ <দຊɾদҪ >

    [5] , & . (2018). /1 1 . In   2018   ( 32 )  (pp. 2J201-2J201). .      [4] , & . (2016). 45 /1 . In 2016 ( 30 ) (pp. 3L3OS16a6-3L3OS16a6). . .
  11. () DEI 6 K K = 6 K K BA=

    ( ( 1 5    P A 2 3 A k R / 04., / 04., T k    new =
  12. 7 . 1 3/ 4/   . . 

    7 7 5 1 3/  
  13. / 3 51 1    4 / 8

    3 51   1. / 8
  14. B /3 4 eo l g 1, 295 g m

    g 1 g 2, r ,, 3, eo b /3 4, eo 5, 5 49 P3 18 P20 21 P41 42 P44 45 P23 39
  15. 1 1 o 4 B . 1 Bloomberg 1 /Bloomberg

    1 1 1 5 l 1 2 0 1 o EBITDA m B /EBITDA PBR m m B g b k e r (k=10,30,60)
  16. 5 2 mB 3 B 5 b 1, 1 4

    le 5 2, , 3, mB B 2 mB g g 4, B g 5, / 4 3 o P3 18 P20 21 P41 42 P44 45 P23 39
  17. . 4 . / 1 . 32 . l .

    4 (MT) l 5 . 4 (MF) l 5 4 3 (MTMF)
  18. / / 2. / 53. / l /1 (MT) l

    /1 (MF) l 1 5 (MTMF) →   (MTMF _TEXT) 4
  19. ) () A K IF MK MK A B= K

    I MK A MK ED _ ) R ( X ( 2    D P 4 5 D P P T / 16., T / 16., k    new MK A F MK 3630 6/ 6
  20. M 5 . /2 X 3 5 6 4 .

    F X 5 4 ET M /2 X ) ( X 5 #"    #"  XGBoost 3 ! #"   … !"# !"$ !"% … !&# !&$ !&% … !'# !'$ !'%   ! 
  21. _ X FE. /4 E E 3 7 E. E

    E 7   TM T … 2018/6/30 31 14 12 10 … 2018/3/30 36 10 14 9 … 2017/12/31 24 14 16 33 … 2017/9/30 54 34 18 23 … 2017/6/30 33 24 37 15 … 2017/3/30 45 33 45 17 … … … … … …  … ↑ 25 E _ X /4 E ) ( Bag of Words(BoW)
  22. ( E X 2 E F X E 4 (

    n XGBoost F/Gradient Boosting Random Forests n 48 n 2 T E 5 / E M 8 2 5 23 _ X E Max_depth = [4,6,8] n_estimators =[50,100,200] XGBoost E 2 . )
  23. 4 2 3 36 5 . 36 / 5 (2014

    6 30 2018 3 30 ) 4 60 ( 3 ) 2018 9 30 3 4,6,8 50,100,200 4 9 / Mean Absolute Percentage Error (MAPE)
  24. M t a b o a bM P l t

    a bM o a b s e 0 3 ). 4 0 ,05/03 0 5545 ), E t A r : M A M eu A gA : ), n M e : !"#$ = 100 ( ) *+, - . /* − /1 /1 . /* ( t /1 ( 3( c b https://mathwords.net/rmsemae t a bM
  25. 5 5 4 MAPE(%) MT MF MTMF MTMF_TEXT MT MF

    MTMF MTMF_TEXT 6.32 3.92 1.65 4.38 11.75 11.41 11.64 17.57 20.30 36.75 35.84 18.42 18.83 19.67 15.47 17.17 1.92 13.17 14.09 12.11 16.98 4.28 3.02 17.20 1.26 3.49 4.23 4.55 12.71 14.55 11.33 11.08 10.63 11.21 5.78 12.19 19.54 5.01 26.77 0.19 28.68 9.51 25.03 17.11 4.00 0.91 3.89 3.83 17.46 3.82 14.19 23.33 2.12 10.27 8.23 1.77 25.69 21.28 23.71 20.41 30.17 30.82 27.91 20.97 24.05 15.63 10.27 10.65 9.47 17.27 8.28 1.55 5.06 7.40 1.08 5.88 13.56 4.95 5.64 8.35 5.37 5.72 3.83 5.77 25.48 17.39 13.09 13.57 5.88 17.78 1.74 2.24 12.81 10.34 10.65 0.27 19.78 7.61 12.80 0.63 7.65 3.36 7.65 3.82 6.32 10.69 10.69 11.63 2.80 3.63 3.30 7.64 11.56 14.80 11.84 14.96 12.27 13.58 17.61 16.68 5 2.98 31.59 31.74 24.96 1.75 13.96 0.00 7.43 10.06 3.11 6.58 11.37 27.71 25.84 21.40 24.48 0.67 6.01 2.33 3.76 42.91 32.03 36.07 31.32
  26. 5 5 4 MAPE(%) MT MF MTMF MTMF_TEXT MT MF

    MTMF MTMF_TEXT 6.32 3.92 1.65 4.38 11.75 11.41 11.64 17.57 20.30 36.75 35.84 18.42 18.83 19.67 15.47 17.17 1.92 13.17 14.09 12.11 16.98 4.28 3.02 17.20 1.26 3.49 4.23 4.55 12.71 14.55 11.33 11.08 10.63 11.21 5.78 12.19 19.54 5.01 26.77 0.19 28.68 9.51 25.03 17.11 4.00 0.91 3.89 3.83 17.46 3.82 14.19 23.33 2.12 10.27 8.23 1.77 25.69 21.28 23.71 20.41 30.17 30.82 27.91 20.97 24.05 15.63 10.27 10.65 9.47 17.27 8.28 1.55 5.06 7.40 1.08 5.88 13.56 4.95 5.64 8.35 5.37 5.72 3.83 5.77 25.48 17.39 13.09 13.57 5.88 17.78 1.74 2.24 12.81 10.34 10.65 0.27 19.78 7.61 12.80 0.63 7.65 3.36 7.65 3.82 6.32 10.69 10.69 11.63 2.80 3.63 3.30 7.64 11.56 14.80 11.84 14.96 12.27 13.58 17.61 16.68 5 2.98 31.59 31.74 24.96 1.75 13.96 0.00 7.43 10.06 3.11 6.58 11.37 27.71 25.84 21.40 24.48 0.67 6.01 2.33 3.76 42.91 32.03 36.07 31.32 MAPE
  27. 3 / 3 4.5 3 3 MT MF MTMF MTMF_TEXT

    13.24 12.85 12.48 11.37 MTMF_TEXT    4 / 3 MAPE(%)
  28. X 3 B B 3 4/5 B 3 B .

    6 7 G      B 3
  29. /X B G B 3.4 B G B 7 57

    /     B
  30. 5 3 4 8 8 . 4 8 8 /

    !$ &. %-  +,* ('  %- %-  /# )" 62.8% 8 31.6% 5.6%
  31. X 3 B B 3 4/5 B 3 B .

    79 G       B 3
  32. B /3 4 eo l g 1, 2 5 g

    m g 1 g 2, r ,, 3, eo b /3 4, eo 5, 0 5 4 P3 18 P20 21 P41 42 P44 45 P23 39
  33. 5 / 5 4 5 4 5 2018 6 30

    2018 9 30 5 2018 6 29 5 1 60 (3 ) 5 9 28 5 1 R 5 . R 5 1 . 1 4 1,000 100 5 .
  34. 4 / 2 4 5 4 MT MF MTMF MTMF_TEXT

    4.61 5.20 4.30 5.89 MTMF_TEXT    / 4MTMF_TEXT4 5 4 . 4 TOPIX 4.89% / . 4 2 4 4 (%)
  35. 5 2 mB 3 B 5 b 1, 1 4

    le 5 2, , 3, mB B 2 mB g g 4, B g 5, / 4 3 o P3 18 P20 21 P41 42 P44 45 P23 39
  36. %% % TO 7 8 A E 7 %% FA

    n XM TP A TO 7 8 n .1. 31 1 . 4 A 4 8 5 n .1. 31 1 1/ % I 8 5 n TO 7 9 n _ 8 XM
  37. avos w t X w _ w a b e

    w t w a w ig ( 2H M 3S PIM 2 (& , G 0 1 CA A P IMG L :P C IMG H ((M ACL IGJ IM PMA I MA C M P MC M JM G I C T P AM A A LIMIMG OO -. -/ 128 ( (& . X l r w fh w a )) 1 4 ) (& k u t _vpyn d i] a -- (& - . (& ) c [ avos my        1 ( (& , a vos fh a 5M (& , )& OO )7)9 ,A, )7)9 ,A, , (& . X fh h a 5M (& . )( OO (6(& (6(& ig