Upgrade to Pro
— share decks privately, control downloads, hide ads and more …
Speaker Deck
Features
Speaker Deck
PRO
Sign in
Sign up for free
Search
Search
俺のNETFLIX season1
Search
h-fkn
April 23, 2020
Technology
0
250
俺のNETFLIX season1
ゆるふわマシーンラーニングvol.3
h-fkn
April 23, 2020
Tweet
Share
More Decks by h-fkn
See All by h-fkn
The advantages and disadvantages of using machine learning with enebular
fkn0839
0
260
ラズパイで写真を撮った話_IoTLT_vol.66_2200812.pdf
fkn0839
0
410
俺のNETFLIX season2 AmazonPersonalize
fkn0839
0
480
ゆるふわマシーンラーニング#2_内容調整中()
fkn0839
0
310
ゆるふわマシーンラーニング「❝ Google AutoML Tablesでお手軽AI ❞と題して話すつもりだったけど、実際に使ったらお手軽()だった件について5分以内で話す」
fkn0839
1
4.1k
データ分析プロセス/AIアプリケーションの基本設計
fkn0839
0
190
DataScienceBOOTCAMP5th_part1
fkn0839
0
2k
G'SACADEMY LAB5th DataScience
fkn0839
0
210
AIアプリ開発に「目的設定」が大切な理由
fkn0839
0
140
Other Decks in Technology
See All in Technology
LLMアプリケーション開発におけるセキュリティリスクと対策 / LLM Application Security
flatt_security
7
1.8k
バイブコーディングと継続的デプロイメント
nwiizo
2
410
組織観点からIAM Identity CenterとIAMの設計を考える
nrinetcom
PRO
1
170
"複雑なデータ処理 × 静的サイト" を両立させる、楽をするRails運用 / A low-effort Rails workflow that combines “Complex Data Processing × Static Sites”
hogelog
3
1.9k
M5製品で作るポン置きセルラー対応カメラ
sayacom
0
140
o11yで育てる、強い内製開発組織
_awache
3
120
What is BigQuery?
aizack_harks
0
130
ACA でMAGI システムを社内で展開しようとした話
mappie_kochi
1
250
[2025-09-30] Databricks Genie を利用した分析基盤とデータモデリングの IVRy の現在地
wxyzzz
0
470
BirdCLEF+2025 Noir 5位解法紹介
myso
0
190
GopherCon Tour 概略
logica0419
2
190
20201008_ファインディ_品質意識を育てる役目は人かAIか___2_.pdf
findy_eventslides
0
120
Featured
See All Featured
Principles of Awesome APIs and How to Build Them.
keavy
127
17k
I Don’t Have Time: Getting Over the Fear to Launch Your Podcast
jcasabona
33
2.5k
Fantastic passwords and where to find them - at NoRuKo
philnash
52
3.4k
Building Applications with DynamoDB
mza
96
6.6k
CSS Pre-Processors: Stylus, Less & Sass
bermonpainter
358
30k
Practical Orchestrator
shlominoach
190
11k
Imperfection Machines: The Place of Print at Facebook
scottboms
269
13k
GitHub's CSS Performance
jonrohan
1032
460k
Typedesign – Prime Four
hannesfritz
42
2.8k
4 Signs Your Business is Dying
shpigford
185
22k
The Invisible Side of Design
smashingmag
301
51k
GraphQLとの向き合い方2022年版
quramy
49
14k
Transcript
2020 ? 1-1 : 96% ! : 0
2020-04-23 Thu #Vol.3 @hidefkn
@hidefkn @AIPM AI 2
2020-04-23 Thu #Vol.3 @hidefkn
NETFLIX 4 1 # ML
season1 5 Google Recommendations AI Google AutoML Tables
Vol.3 2020-04-23 Thu NETFLIX
NETFLIX ! 7
! 8
NETFLIX NETFLIX 9 • Ͳ͏࣮͍ͬͯͯ͠Δͷ͔ʁ • ͲΜͳσʔλੳΛ͍ͯ͠Δͷ͔ʁ • ԶͬͯΈΑ͏ʔͬͱʂ
Vol.3 2020-04-23 Thu
1,200 1955 2019 11 1,278 616
1 1.4 Film Exhibition Yearbook 2018 12
Film Exhibition Yearbook 2018 13 1,200 1.4 1 3.3 1
2 6 ! ! ! ! ! !
Vol.3 2020-04-23 Thu
15 ! よよ GW 5 ! よよ !よよ NETFLIX
Vol.3 2020-04-23 Thu NETFLIX
NETFLIX 18 εϥμϯݟͯ Δؾग़ͯͨ͠
NETFLIX 19 ϫΠͷࢹௌཤྺ
NETFLIX MovieLens 20 • MovieLens GroupLens • NETFLIX https://movielens.org/
2018 9 9 1MB 21 ml-latest-small.zip ratings.csv movies.csv tags.csv links.csv
• 600 • 9,000 • • • 100,000 ;ʙΜʢʁʣ
2018 9 9 1MB 22 ml-latest-small.zip ratings.csv movies.csv tags.csv links.csv
• 2 • • 1 ! ! ΏΔ͘Γ͍ͨ
ratings.csv 23 • userId • timestamp • movieId • rating
movieId movies.csv 24
userId 4 movies.csv 25 ͦΖͦΖ࣌ؒ͋ΕͳͷͰɺ ૣ͍ͱ͜ΖɺΫϥυ.-ʹΈ͍ͨ
Vol.3 2020-04-23 Thu Google RecommendationsAI
Google Recommendations AI 27 Recommendations AI ML ML
Google Recommendations AI 28
Recommendations AI ! 30 Recommendation AI Prediction API Show recommendations
at customer touchpoints 1 INGET YOUR DATA データを取り込む 2 CUSTOMIZE YOUR RECOMMENDATIONS 推薦システムをカスタムする 3 DELIVER RECOMMENDATIONS ANYWHERE IN THE CUSTOMER JOURNY 推薦システムをユーザー導線に提供する
31 ! ( )
32 Google Cloud A/B
Google Cloud 33 ⥯ך (PPHMF$MPVEفٗآؙؑزإحز،حفׅ
Google Cloud 34
Google Cloud 35 ۀΞΧϯτ͕ͳ͍ͱ ༡ΜͰΈΔͷ͕ͦ͠͏ શવΏΔ͘ͳ͍ɻ ΨνײΛײ͡Δɻ
Google 36
Vol.3 2020-04-23 Thu Excel AutoML Tables
38 Good 1, Bad 0 A A A A E
A 1 1 0 1 1 1 (100%) B 1 1 0 1 0 0.61 61.24%) C 0 0 1 0 0 -1 (-100%)
userId 1 266 ! 39 εϓϨουγʔτ ·͡ΏΔͬͱͰ͖ΔΘ͊ʙ
40 ݟ͍ͯΔөը͕ҰॹͰɺͭɺධՁ͕ࣅ͍ͯΔ ˠ͋ͳͨͱΈ͕ࣅ͍ͯΔΑʂͬͯ͜ͱͶ
41 ΞΫγϣϯܥͷөըΛߴධՁͯ͠ΔݟͯΔϢʔβʔ ΞΫγϣϯܥͷөըΛߴධՁ͍ͯ͠ΔϢʔβʔͱ૬͍ؔͯ͠Δʢࣅ͍ͯΔʣ ˠ ͦΓΌͦ͏Ͷ ˠ 1 !
AutoML Tables 42
AutoML Tables 43 • 500 • 1000 • 600 •
1000 ͔ͩΒͦ͜ɺ3FDPNNFOEBUJPOT"*͕͋ΔͷͶʜ ͦͦతҧ͏͠ͳ͊
44
45 ɺ"VUP.- 5BCMFT Ͱ༡Μ͚ͩͩͰ͢ɻ
46
2020 ? 1-1 : 96% ! : 47
Season1