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

ロジスティック回帰 Part 1 - 基礎編

19fc8f6113c5c3d86e6176362ff29479?s=47 Kan Nishida
PRO
September 19, 2019
Science
0
64

ロジスティック回帰 Part 1 - 基礎編

19fc8f6113c5c3d86e6176362ff29479?s=128

Kan Nishida
PRO

September 19, 2019
Tweet

More Decks by Kan Nishida

See All by Kan Nishida
19fc8f6113c5c3d86e6176362ff29479?s=48 kanaugust
PRO
0
52
19fc8f6113c5c3d86e6176362ff29479?s=48 kanaugust
PRO
0
460
19fc8f6113c5c3d86e6176362ff29479?s=48 kanaugust
PRO
0
300
19fc8f6113c5c3d86e6176362ff29479?s=48 kanaugust
PRO
0
180
19fc8f6113c5c3d86e6176362ff29479?s=48 kanaugust
PRO
0
190
19fc8f6113c5c3d86e6176362ff29479?s=48 kanaugust
PRO
0
90
19fc8f6113c5c3d86e6176362ff29479?s=48 kanaugust
PRO
0
62
19fc8f6113c5c3d86e6176362ff29479?s=48 kanaugust
PRO
0
67
19fc8f6113c5c3d86e6176362ff29479?s=48 kanaugust
PRO
0
120

Other Decks in Science

See All in Science
6402413ab839ccca6cce08c8bbd51827?s=48 terensarnic
0
300
6734b737614a599f9b899f190e413d2d?s=48 deepflow
0
250
A2cac4b3dcb2bc0b87917ddc034ef708?s=48 sansandsoc
0
630
7975b9fd58c8945ae1c6b38747de7f28?s=48 line_developers_tw2
0
2.6k
40044819b1dc2904638f689281150cac?s=48 seansaito
0
960
7b31e2e5598a145a2b3d97323b26a514?s=48 harunashi
0
170
E55f04d00716de969f69dd3c4475dc92?s=48 calpis10000
0
270
4fe5193960c3289769e3913281560dc4?s=48 taka88
1
160
948966d9c690e72faba4fd76e1858c56?s=48 tagtag
0
120
99e9f426c1e04a86aecb70f65f4d05ef?s=48 taketosano
0
340
654d48d6c1c10c50c160954ba31207a2?s=48 rabernat
0
870
17903fabbd6615c7c178c3eff89bd03e?s=48 mitti1210
2
810

Featured

See All Featured
719435d98d452de7ac367c828266cf01?s=48 erikaheidi
29
5.1k
A9179349dd2bdc67f377719f56d85656?s=48 garrettdimon
293
110k
651dcfe41723207fbb0aa044a4f7c07f?s=48 dotmariusz
99
5.7k
F57e2136eb9895eb95ff5dc8a1c02677?s=48 zakiwarfel
91
3.6k
E6c6e133e74c3b83f04d2861deaa1c20?s=48 searls
207
38k
F29327647a9cff5c69618bae420792ea?s=48 tenderlove
58
3.8k
Af6a12710770ad9a7cfd08c97efd40d3?s=48 pedronauck
655
120k
B47b288784fda77a94aeb234ca743c24?s=48 keavy
111
15k
282d599b414022eba5096510f675b6e2?s=48 chrislema
174
14k
7c8469ee8c9e594c65c59b919626c08d?s=48 scottboms
253
11k
20bfe76b3d6105641f879fe45cfc9272?s=48 bkeepers
PRO
324
54k
5b9fe87ec1faa67a4599782930f45ec9?s=48 sstephenson
149
12k

Transcript

  1. ϩδεςΟοΫճؼ Part 1 جૅฤ Exploratory Seminar #19

  2. EXPLORATORY

  3. 3 εϐʔΧʔ ੢ా צҰ࿠ CEO EXPLORATORY ུྺ 2016೥ɺσʔλαΠΤϯεͷຽओԽͷͨΊɺExploratory, Inc Λ

    ্ཱͪ͛Δɻ Exploratory, Inc.ͰCEOΛ຿ΊΔ͔ͨΘΒɺσʔλαΠΤϯεɾ ϒʔτΩϟϯϓɾτϨʔχϯάͳͲΛ௨ͯ͠γϦίϯόϨʔͰ ߦΘΕ͍ͯΔ࠷ઌ୺ͷσʔλαΠΤϯεͷීٴͱڭҭʹऔΓ૊ Ήɻ ถΦϥΫϧຊࣾͰɺ16೥ʹΘͨΓσʔλαΠΤϯεͷ։ൃνʔ ϜΛ཰͍ɺػցֶशɺϏοάɾσʔλɺϏδωεɾΠϯςϦδΣ ϯεɺσʔλϕʔεʹؔ͢Δ਺ଟ͘ͷ੡඼ΛੈʹૹΓग़ͨ͠ɻ @KanAugust

  4. Vision ΑΓΑ͍ҙࢥܾఆΛ͢ΔͨΊʹ σʔλΛ࢖͏͜ͱ͕౰ͨΓલʹͳΔ

  5. Mission σʔλαΠΤϯεͷຽओԽ

  6. 6 ୈ̏ͷ೾ σʔλαΠΤϯεɺAIɺػցֶश͸౷ܭֶऀɺ։ൃऀͷͨΊ͚ͩͷ΋ͷͰ͸͋Γ·ͤΜɻ σʔλʹڵຯͷ͋ΔਓͳΒ୭΋͕ੈքͰ࠷ઌ୺ͷΞϧΰϦζϜΛ࢖ͬͯ ϏδωεσʔλΛ؆୯ʹ෼ੳͰ͖Δ΂͖Ͱ͢ɻ Exploratory͕ͦ͏ͨ͠ੈքΛՄೳʹ͠·͢ɻ

  7. ୈ1ͷ೾ ୈ̎ͷ೾ ୈ̏ͷ೾ ϓϥΠϕʔτ(ߴ͍/ݹ͍) Φʔϓϯɾιʔε(ແྉ/࠷ઌ୺) UI & ϓϩάϥϛϯά ϓϩάϥϛϯά 2016

    2000 1976 ϚωλΠθʔγϣϯ ίϞσΟςΟԽ ຽओԽ ౷ܭֶऀ σʔλαΠΤϯςΟετ Exploratory ΞϧΰϦζϜ Ϣʔβʔɾ ମݧ πʔϧ Φʔϓϯɾιʔε(ແྉ/࠷ઌ୺) UI & ࣗಈԽ ϏδωεɾϢʔβʔ ςʔϚ σʔλαΠΤϯεͷຽओԽ

  8. 質問 ExploratoryɹϞμϯˍγϯϓϧ UI 伝える データアクセス データ ラングリング 可視化 アナリティクス 統計/機械学習

  9. ϩδεςΟοΫճؼ Part 1 جૅฤ Exploratory Seminar #19

  10. 質問 伝える データアクセス データ ラングリング 可視化 アナリティクス 統計/機械学習

  11. ઢܗճؼͷ෮श

  12. USͷ੺ͪΌΜσʔλ

  13. ໰୊ ෕਌ͷ೥ྸ͸͍͔ͭ͘ɺ฼਌ͷ೥ྸΛ΋ͱʹ༧ଌ͍ͨ͠ɻ

  14. ෕਌ͷ೥ྸ vs. ฼਌ͷ೥ྸ

  15. ઢܗճؼ

  16. Father_Age = 0.866 * Mother_Age + 6.28 ੾ย ܎਺ʢ܏͖ʣ

  17. ڵຯͷର৅ ਺஋ ΧςΰϦʔ/ೋ߲ 17 ΧςΰϦʔ/ଟ߲

  18. ڵຯͷର৅ ਺஋ ΧςΰϦʔ/ೋ߲ 18 ΧςΰϦʔ/ଟ߲

  19. ͜ͷϢʔβʔ͸ίϯόʔτ͢Δ͔ʁ ͜ͷऔҾ͸ෆਖ਼͔ʁ ͜ͷैۀһ͸΍ΊΔ͔ʁ ͜ͷ੺ͪΌΜ͸ະख़ࣇͰੜ·ΕΔ͔ʁ ೋ߲ͷ࣭໰

  20. 20 ਺஋ͷ໰୊Λೋ߲ͷ໰୊΁

  21. 21 ੺ͪΌΜͷ೛৷ظؒ

  22. 22 premature = gestation week < 37 ੜ·Εͯ͘Δ੺ͪΌΜ͸ະख़ࣇ͔ʁ TRUE FALSE

  23. Numeric Binary ਺஋ͷ໰୊Λೋ߲෼ྨͷ໰୊΁

  24. ໰୊ ෕਌͕35ࡀΑΓ্͔Ͳ͏͔ɺ฼਌ͷ೥ྸΛ΋ͱʹ༧ଌ͍ͨ͠ɻ

  25. ճؼͷΞϧΰϦζϜΛ࢖ͬͯೋ߲ͷ໰୊ΛղܾͰ͖ͳ͍͔ʁ

  26. YES YOU CAN!

  27. ͔͠͠ɺͪΐͬͱ޻෉͕ඞཁͰɺੲͷਓ͸ۤ͠Μͩɻ

  28. ͓͔͛Ͱɺࢲୡ͸ϩδεςΟοΫճؼͱͯ͠࢖͏͚ͩɻ

  29. ͲΜͳ͜ͱ΍ͬͯΔͷ͔ͪΐͬͱ೷͍ͯΈ·͠ΐ͏ɻ

  30. 1. TRUE/FALSEΛ̌/̍ͱ͍͏਺஋ʹͯ֬͠཰Λ༧ଌ͢Δ໰୊ʹ͢Δ ϩδεςΟοΫճؼͷ࢓૊Έ

  31. 1. TRUE/FALSEΛ̌/̍ͱ͍͏਺஋ʹͯ֬͠཰Λ༧ଌ͢Δ໰୊ʹ͢Δ 2. ͦͷ··ͩͱ͍͔ͭ͘໰୊͕͋ΔͷͰσʔλΛՃ޻͢Δ ϩδεςΟοΫճؼͷ࢓૊Έ

  32. 1. TRUE/FALSEΛ̌/̍ͱ͍͏਺஋ʹͯ֬͠཰Λ༧ଌ͢Δ໰୊ʹ͢Δ 2. ͦͷ··ͩͱ͍͔ͭ͘໰୊͕͋ΔͷͰσʔλΛՃ޻͢Δ 3. ༧ଌ͞ΕΔ֬཰Λ΋ͱʹTRUE͔FALSEͷϥϕϧ෇͚Λͯ͠Ξ΢τϓοτ͢Δ ϩδεςΟοΫճؼͷ࢓૊Έ

  33. 1. TRUE/FALSEΛ̌/̍ͱ͍͏਺஋ʹͯ֬͠཰Λ༧ଌ͢Δ໰୊ʹ͢Δ 2. ͦͷ··ͩͱ͍͔ͭ͘໰୊͕͋ΔͷͰσʔλΛՃ޻͢Δ 3. ༧ଌ͞ΕΔ֬཰Λ΋ͱʹTRUE͔FALSEͷϥϕϧ෇͚Λͯ͠Ξ΢τϓοτ͢Δ 1. ֬཰ 2. Φοζ

    3. ϩάɾΦοζ 4. ֬཰ ϩδεςΟοΫճؼͷ࢓૊Έ

  34. Step by Step

  35. 1. TRUE/FALSEΛ̌/̍ͱ͍͏਺஋ʹͯ֬͠཰Λ༧ଌ͢Δ໰୊ʹ͢Δ 2. ͦͷ··ͩͱ͍͔ͭ͘໰୊͕͋ΔͷͰσʔλΛՃ޻͢Δ 3. ༧ଌ͞ΕΔ֬཰Λ΋ͱʹTRUE͔FALSEͷϥϕϧ෇͚Λͯ͠Ξ΢τϓοτ͢Δ 1. ֬཰ 2. Φοζ

    3. ϩάɾΦοζ 4. ֬཰ ϩδεςΟοΫճؼͷ࢓૊Έ

  36. Binary Numeric TRUE or FALSE 1 or 0

  37. None
  38. None

  39. ֬཰ 100% 0% ฼਌ͷ೥ྸ

  40. ਺஋ͷ༧ଌͳͷͰઢܗճؼʹ͔͚ͯΈΑ͏ʂ

  41. ઢܗճؼͷϞσϧ

  42. P(Father > 35) = a * Mother_Age + b P():

    ֬཰ΛٻΊΔؔ਺
  43. None

  44. Pr(Father > 35) = 0.039 * Mother_Age -0.85

  45. ઢܗճؼͷϞσϧ Pr(Father > 35) = 0.039 * Mother_Age -0.85

  46. ͓฼͞Μ͕35ࡀͷͱ͖ͷ֬཰͸ʁ

  47. Pr(Father > 35) = 0.039 * 35 - 0.85 =

    0.515 51.5% ͷ֬཰Ͱ෕਌͸35ࡀΑΓ্ɻ Pr(Father > 35) = 0.039 * Mother_Age -0.85 ฼਌ͷ೥ྸ: 35

  48. 35ࡀ 51.5%

  49. ͓฼͞Μ͕20ࡀͷͱ͖ͷ֬཰͸ʁ

  50. Pr(Father > 35) = 0.039 * 20 - 0.85 =

    -0.07 Pr(Father > 35) = 0.039 * Mother_Age -0.85 ϚΠφε 7% ͷ֬཰Ͱ෕਌͸35ࡀΑΓ্ɻ ฼਌ͷ೥ྸ: 20

  51. 20ࡀ -7%

  52. ϚΠφεͷ֬཰ʁʁʁ

  53. ϚΠφεͷ֬཰͸ڹ͖͸͓΋͠Ζ͍͕ɺ ࣮ࡍʹ͸શ͘ҙຯΛͳ͞ͳ͍ɻ

  54. ͜ͷลΓʹ͘Δਓͨͪͷઆ໌͕͏·͘Ͱ͖ͳ͍ɻ

  55. Ͱ΋ɺσʔλ͸͜ͷลʹ΋͔ͬ͠Γ͋ΔͷͰɺͪΌΜͱઆ໌Ͱ͖ΔϞ σϧ͕΄͍͠ɻ

  56. ઢܗճؼ͸0͔Β1·Ͱͷ஋͔͠औΒͳ͍֬཰Λ༧ଌ͢ΔͨΊʹ ͸͋·Γద͍ͯ͠ͳ͍ɻ

  57. ্ݶɺԼݶΛ͚ͭΔͷ͸Ͳ͏ͩʁ 0% 100%

  58. ͱ͍͏͜ͱ͸ɺ฼਌͕21ࡀΑΓए͔ͬͨΒ0ˋͱ͍͏͜ͱɻ 0% 100%

  59. 0% 100% ͔͠͠ɺ࣮ࡍʹ͸1(෕਌ͷ೥ྸ͕35ΑΓ্ʣͷਓୡ΋͍Δɻ

  60. 0% 100% ͓͍͠ɺ΋͏Ұ޻෉ඞཁʂ

  61. σʔλͷม׵ʂ

  62. 1. TRUE/FALSEΛ̌/̍ͱ͍͏਺஋ʹͯ֬͠཰Λ༧ଌ͢Δ໰୊ʹ͢Δ 2. ͦͷ··ͩͱ͍͔ͭ͘໰୊͕͋ΔͷͰσʔλΛՃ޻͢Δ 3. ༧ଌ͞ΕΔ֬཰Λ΋ͱʹTRUE͔FALSEͷϥϕϧ෇͚Λͯ͠Ξ΢τϓοτ͢Δ 1. ֬཰ 2. Φοζ

    3. ϩάɾΦοζ 4. ֬཰ ϩδεςΟοΫճؼͷ࢓૊Έ

  63. ͲΜͳม׵͕Ͱ͖Δ͔ʁ ֬཰ͩͱ0͔Β1ͷؒͱ͍͏਺஋ͷൣғʹ੍ݶ͕͋Δͷ͕໰୊ͩɻ ճؼͷϞσϧΛ࢖͏ʹ͸࿈ଓ஍Ͱ஋ʹ੍ݶ͕ͳ͍΄͏͕͍͍ɻ [- Infinity - infinity] ੍ݶͷͳ͍࿈ଓ஋ [0 -

    1] range ੍ݶͷ͋Δൣғ

  64. Logit (Logistic Unit) ϩδοτؔ਺

  65. Logit (Logistic Unit) Log of Odds ϩάɾΦοζ

  66. 66 Φοζ(Odds) ى͖͏Δೋͭͷ݁Ռͷ֬཰ͷൺ

  67. 67 Φοζ Φοζ = TRUEͷ֬཰ / FALSEͷ֬཰

  68. 68 ෕਌͕35ࡀΑΓ্Ͱ͋ΔΦοζ Φοζ = TRUEͷ֬཰ / FALSEͷ֬཰ ෕਌͕35ࡀΑΓ্ͷ֬཰͕10% ෕਌͕35ࡀΑΓ্ͷ֬཰͕90% 0.1111…

    = 10 / 90

  69. ֬཰ vs. Φοζ P(Father > 35) = 0.2 P(TRUE) =

    P(Father > 35) = 0.2 P(FALSE): 1 - P(Father > 35) = 0.8 Φοζ = P(TRUE) / P(FALSE) = 0.2 / 0.8 = 0.25

  70. ֬཰ vs. Φοζ P(Father > 35) = 0.75 P(TRUE) =

    P(Father > 35) = 0.75 P(FALSE): 1 - P(Father > 35) = 0.25 Φοζ = P(TRUE) / P(FALSE) = 0.75 / 0.25 = 3

  71. Pr(Father > 35) = 0 0 / (1 - 0)

    = 0 ֬཰ Φοζ Pr(Father > 35) = 0.5 0.5 / (1 - 0.5) = 1 Pr(Father > 35) = 0.9 0.9 / (1 - 0.9) = 9 Pr(Father > 35) = 0.999 0.999 / (1 - 0.999) = 999 Pr(Father > 35) = 1 1 / (1 - 1) = ແݶ ม׵

  72. Probability can only range from 0 to 1, Odds can

    be 0 up to any positive number. But, we still have a problem. We want the variable that can range from any negative number to any positive number.

  73. ֬཰ 0 1 Φοζ 0 1 ແݶ

  74. ֬཰ 0 1 Φοζ 0 1 ແݶ ແݶ -ແݶ 0

    ཧ૝

  75. ϩάɾΦοζ log( P(y) 1 - P(y) ) ΦοζʹϩάΛ͔͚Δ

  76. None

  77. ֬཰ 0 1 Φοζ 0 1 ແݶ ແݶ -ແݶ 0

    ϩάɾΦοζ log( Odds( P(y) )) Odds( P(y) )

  78. ϩδοτؔ਺ log( Odds( P(y) )) = Logit(P(y)) ֬཰ΛϩάɾΦοζʹม׵ͯ͘͠ΕΔؔ਺

  79. • ֬཰͸0͔Β1ͷؒͷൣғͷ஋͚ͩΛؚΉɻ • Φοζ͸0͔Βϓϥεແݶେʹଓ͘਺஋ͷൣғͷ஋͚ͩΛؚΉɻ • ϩάɾΦοζ͸ϚΠφεແݶେ͔ΒϓϥεແݶେͷؒͷͲΜͳ஋ Ͱ΋औΓ͏Δɻ

  80. ϩάɾΦοζ͸ͲΜͳ਺஋Ͱ΋ͱΕΔɻ ͱ͍͏͜ͱ͸ɺ࿈ଓ਺஋Λ༧ଌ͢ΔͨΊͷճؼͷΞϧΰϦζϜ ͕࢖͑Δʂ

  81. ϩδεςΟοΫճؼ खݩʹ͋Δ༧ଌม਺Λݩʹͯ͠ɺڵຯͷର৅Ͱ͋Δೋ߲ม਺ͷ ϩάɾΦοζΛ༧ଌ͢ΔͨΊͷϞσϧΛ࡞ΔͨΊͷΞϧΰϦζ Ϝɻ

  82. ઢܗճؼ ϩδεςΟοΫճؼ Logit( P(Father > 35) ) = a *

    Mother_Age + b Father_Age = a * Mother_Age + b

  83. ϩδεςΟοΫճؼΛྲྀ͠ݟͯΈΔʂ

  84. None
  85. None

  86. Logit( Pr(Father > 35) ) = 0.029 * Mother_Age -

    0.1012

  87. ฼਌͕40ࡀͷ࣌ɺ෕਌͕35ࡀҎ্Ͱ͋Δ֬཰͸ʁ

  88. Logit( P(Father > 35) ) = 0.29 * Mother_Age -

    10.12 Logit( P(Father > 35) ) = 0.29 * 40 - 10.12 = 1.48 ෕਌͕35ࡀҎ্Ͱ͋ΔϩάɾΦοζ͸1.48ɻ ???

  89. ϩάɾΦοζͷཧղ͸ࢲͷ௚ײΛ௒͍͑ͯΔɻɻɻ

  90. ϩάɾΦοζΛͻͬ͘Γฦͯ֬͠཰ ʹ໭͍ͨ͠ɻ

  91. 1. TRUE/FALSEΛ̌/̍ͱ͍͏਺஋ʹͯ֬͠཰Λ༧ଌ͢Δ໰୊ʹ͢Δ 2. ͦͷ··ͩͱ͍͔ͭ͘໰୊͕͋ΔͷͰσʔλΛՃ޻͢Δ 3. ༧ଌ͞ΕΔ֬཰Λ΋ͱʹTRUE͔FALSEͷϥϕϧ෇͚Λͯ͠Ξ΢τϓοτ͢Δ 1. ֬཰ 2. Φοζ

    3. ϩάɾΦοζ 4. ֬཰ ϩδεςΟοΫճؼͷ࢓૊Έ

  92. Logit(P(Father > 35)) = 0.03 * Mother_Age - 0.85 P(Father

    > 35) = Logit (0.03 * Mother_Age - 0.85) -1 ٯؔ਺

  93. Logit(Pr(Father > 35)) = 0.03 * Mother_Age - 0.85 Pr(Father

    > 35) = (0.03 * Mother_Age - 0.85) Logistic

  94. P(Father > 35) = Logistic(0.29 * Mother_Age - 10.12)

  95. ฼਌͕40ࡀͷ࣌ɺ෕਌͕35ࡀҎ্Ͱ͋Δ֬཰͸ʁ

  96. P(Father > 35) = Logistic(0.29 * Mother_Age - 10.12) P(Father

    > 35) = Logistic(0.29 * 40 - 10.12) = Logistic(1.48) = 0.8145 ฼਌͕40ࡀͷ࣌ɺ෕਌͕35ࡀҎ্Ͱ͋Δ֬཰͸81%ɻ

  97. ฼਌͕20ࡀͷ࣌ɺ෕਌͕35ࡀҎ্Ͱ͋Δ֬཰͸ʁ

  98. P(Father > 35) = Logistic(0.29 * Mother_Age - 10.12) P(Father

    > 35) = Logistic(0.29 * 20 - 10.12) = Logistic(-4.32) = 0.01312 ฼਌͕20ࡀͷ࣌ɺ෕਌͕35ࡀҎ্Ͱ͋Δ֬཰͸1.3%ɻ

  99. ͜ͷϩδεςΟοΫճؼͷϞσϧ͸ͲΜͳ͔Μ͡ͷ΋ͷʁ

  100. 100 ઢܗճؼͷϞσϧ vs ϩδεςΟοΫճؼͷϞσϧ

  101. ઢܗճؼͷ৔߹ΛݟͯΈΔɻ

  102. 102 Father_Age = a * Mother_Age + b ܎਺ʢ܏͖ʣ ੾ย

    ઢܗճؼͷϞσϧʢܭࢉࣜʣ

  103. 103 ܎਺ʢ܏͖ʣ ੾ย

  104. 104 Father_Age = 0.87 * Mother_Age + 6.28 ܎਺ʢ܏͖ʣ ੾ย

    ઢܗճؼͷϞσϧʢܭࢉࣜʣ
  105. None

  106. ෕਌ͷ೥ྸ ฼਌ͷ೥ྸ

  107. ෕਌ͷ೥ྸ ฼਌ͷ೥ྸ

  108. How about this Logistic Regression Model?

  109. 109 ෕਌͕35Ҏ্ͷ֬཰ = logistic(a * Mother_Age + b) ܎਺ʢ܏͖ʣ ੾ย

    ܎਺ͱ੾ยΛௐઅ͢Δ͜ͱͰ࣮σʔλͱ Ϛον͢ΔΑ͏ͳۂઢ͕ඳ͚Δɻ

  110. 110 ϩδεςΟοΫճؼͷϞσϧ

  111. 111 ෕਌͕35Ҏ্ͷ֬཰ = logistic(0.29 * Mother_Age - 10.12) ੾ย ܎਺

  112. None

  113. ֬཰ (Father > 35) ฼਌ͷ೥ྸ

  114. ϩδεςΟοΫۂઢ

  115. ֬཰ (Father > 35) ฼਌ͷ೥ྸ

  116. ࣮σʔλ Ϟσϧ (ϩδεςΟοΫۂઢ) ΋ͱͷσʔλͷฏۉΛऔΔͱɻɻɻ

  117. ͱ͜ΖͰɺຊ౰ʹ΍Γ͔ͨͬͨͷ͸ೋ߲ͷ֬཰Ͱͳ ͯ͘ɺ஋ʢTRUE/FALSEʣͷ༧ଌɻ ฼਌ͷ೥ྸΛ΋ͱʹɺ෕਌͕̏̑ࡀҎ্ͳͷ͔Ͳ͏ ͔Λ஌Γ͍ͨɻ

  118. ͱ͜ΖͰɺຊ౰ʹ΍Γ͔ͨͬͨͷ͸ೋ߲ͷ֬཰Ͱͳ ͯ͘ɺ஋ʢTRUE/FALSEʣͷ༧ଌɻ ฼਌ͷ೥ྸΛ΋ͱʹɺ෕਌͕̏̑ࡀҎ্ͳͷ͔Ͳ͏ ͔Λ஌Γ͍ͨɻ TRUE or FALSE ?

  119. 1. TRUE/FALSEΛ̌/̍ͱ͍͏਺஋ʹͯ֬͠཰Λ༧ଌ͢Δ໰୊ʹ͢Δ 2. ͦͷ··ͩͱ͍͔ͭ͘໰୊͕͋ΔͷͰσʔλΛՃ޻͢Δ 3. ༧ଌ͞ΕΔ֬཰Λ΋ͱʹTRUE͔FALSEͷϥϕϧ෇͚Λͯ͠Ξ΢τϓοτ͢Δ 1. ֬཰ 2. Φοζ

    3. ϩάɾΦοζ 4. ֬཰ ϩδεςΟοΫճؼͷ࢓૊Έ

  120. 120 100% 0% 30 35 40 ฼਌ͷ೥ྸ ෕਌ͷ೥ྸ͕ 35ࡀҎ্Ͱ͋Δ ֬཰

    ֬཰Ͱ͸ͳ͘TRUE͔FALSEʹ෼ྨ͍ͨ͠ɻ

  121. 121 30 35 40 ฼਌ͷ೥ྸ TRUE FALSE ෕਌ͷ೥ྸ͕ 35ࡀҎ্ 100%

    0% ֬཰Ͱ͸ͳ͘TRUE͔FALSEʹ෼ྨ͍ͨ͠ɻ

  122. 122 30 35 40 ฼਌ͷ೥ྸ 50% ෕਌ͷ೥ྸ͕ 35ࡀҎ্ TRUE FALSE

    ͖͍͠஋Λઃ͚Δɻྫ͑͹ɺ֬཰͕50%Λڥʹ͢Δɻ

  123. 123 30 35 40 ฼਌ͷ೥ྸ 50% ෕਌ͷ೥ྸ͕ 35ࡀҎ্ TRUE FALSE

    TRUE FALSE ͜ͷۂઢΛ࢖͏͜ͱͰ෼ྨ͢Δ͜ͱ͕Ͱ͖ΔΑ͏ʹͳͬͨɻ

  124. 124 30 35 40 ฼਌ͷ೥ྸ 50% ෕਌ͷ೥ྸ͕ 35ࡀҎ্ TRUE FALSE

    TRUE FALSE ͜ͷۂઢΛ࢖͏͜ͱͰ෼ྨ͢Δ͜ͱ͕Ͱ͖ΔΑ͏ʹͳͬͨɻ

  125. 125 30 35 40 ฼਌ͷ೥ྸ 50% ෕਌ͷ೥ྸ͕ 35ࡀҎ্ TRUE FALSE

    TRUE FALSE ࣮ࡍʹ͸ɺ෕਌ͷ೥ྸ͕35ࡀҎ্ͩͬͨσʔλʹ౰ͯ͸ΊͯΈΔ

  126. 126 30 35 40 ฼਌ͷ೥ྸ 50% ෕਌ͷ೥ྸ͕ 35ࡀҎ্ TRUE FALSE

    TRUE FALSE ౰͍ͨͬͯΔ ౰͍ͨͬͯͳ͍

  127. 127 30 35 40 ฼਌ͷ೥ྸ 50% ෕਌ͷ೥ྸ͕ 35ࡀҎ্ TRUE FALSE

    TRUE FALSE ࣮ࡍʹ͸ɺ෕਌ͷ೥ྸ͕35ࡀҎ্͡Όͳ͔ͬͨσʔλʹ౰ͯ͸ΊͯΈΔ

  128. 128 30 35 40 50% ෕਌ͷ೥ྸ͕ 35ࡀҎ্ TRUE FALSE TRUE

    FALSE ౰͍ͨͬͯΔ ౰͍ͨͬͯͳ͍

  129. 129 30 35 40 50% ෕਌ͷ೥ྸ͕ 35ࡀҎ্ TRUE FALSE TRUE

    FALSE ౰͍ͨͬͯͳ͍ ౰͍ͨͬͯͳ͍

  130. 130 = 12/14 = 0.857 (85.7%) ͜ͷۂઢϞσϧΛ࢖͏ͱɺ 85.7%ͷਫ਼౓Ͱ༧ଌͰ͖Δɻ

  131. 131 ༧ଌม਺͕1͚ͭͩͷ৔߹ΛΈ͖ͯͨɻ

  132. 132 ༧ଌม਺͕ෳ਺ͷ৔߹͸ʁ

  133. 133 ෕਌͕35Ҏ্ͷ֬཰ = logistic(0.3 * Mother_Age - 10) ੾ย ܎਺

    ༧ଌม਺͕Mother_Age͚ͩͷ৔߹

  134. 134 ෕਌͕35Ҏ্ͷ֬཰ = logistic(0.3 * Mother_Age + 1.2 * Mother_Japanese

    - 10) ༧ଌม਺͕Mother_AgeͱMother_Japaneseͷ ̎ͭͷ৔߹

  135. Q & A

  136. None

  137. • ϓϩάϥϛϯάͳ͠ RݴޠͷUIͰ͋ΔExploratoryΛ෼ੳπʔϧͱͯ͠࢖༻͢ΔͨΊडߨத͸ɺϏδωεͷ໰୊ Λղܾ͢ΔͨΊʹඞཁͳσʔλαΠΤϯεͷख๏ͷशಘʹ100ˋूதͰ͖Δ • πʔϧͷ࢖͍ํͰ͸ͳ͘ɺ෼ੳख๏ͷशಘ ݱ৔Ͱ࢖͑Δ෼ੳख๏ΛάϧʔϓԋशΛ௨࣮ͯ͠ࡍʹखΛಈ͔͠ͳ͕Βɺ਎ʹ͚ͭͯߦ͘ ͜ͱ͕Ͱ͖Δɻ • ࢥߟྗͱεΩϧͷशಘ

    σʔλαΠΤϯεͷεΩϧशಘ͚ͩͰͳ͘ɺσʔλ෼ੳʹඞཁͳࢥߟྗ΋शಘͰ͖Δ ಛ௃

  138. ࿈བྷઌ ϝʔϧ kan@exploratory.io ΢ΣϒαΠτ https://ja.exploratory.io ϒʔτΩϟϯϓɾτϨʔχϯά https://ja.exploratory.io/training-jp Twitter @KanAugust

  139. EXPLORATORY