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
Amazon Personalizeではじめるレコメンドサービス
Search
tandfy
November 29, 2019
Technology
0
1.2k
Amazon Personalizeではじめるレコメンドサービス
HIGOBASHI.AWS 第12回 活用編のセッション「Amazon Personalizeではじめるレコメンドサービス」の資料です。
tandfy
November 29, 2019
Tweet
Share
More Decks by tandfy
See All by tandfy
アプリから集まるイベントデータのリアルタイム処理入門
tandfy
0
1.6k
Amazon SageMakerの最新アップデートの紹介
tandfy
1
910
DeepRacerで学ぶ機械学習 1.1
tandfy
0
1k
DeepRacerで始める機械学習
tandfy
1
1.7k
DeepRacerでまなぶ強化学習
tandfy
1
1.4k
Amazon SageMakerではじめる物体検出
tandfy
1
1.1k
Other Decks in Technology
See All in Technology
AWSに革命を起こすかもしれない新サービス・アップデートについてのお話
yama3133
0
510
フルカイテン株式会社 エンジニア向け採用資料
fullkaiten
0
9.9k
AWSインフルエンサーへの道 / load of AWS Influencer
whisaiyo
0
230
Authlete で実装する MCP OAuth 認可サーバー #CIMD の実装を添えて
watahani
0
200
TED_modeki_共創ラボ_20251203.pdf
iotcomjpadmin
0
150
AgentCoreとStrandsで社内d払いナレッジボットを作った話
motojimayu
1
990
AWSの新機能をフル活用した「re:Inventエージェント」開発秘話
minorun365
2
470
2025年のデザインシステムとAI 活用を振り返る
leveragestech
0
330
AIBuildersDay_track_A_iidaxs
iidaxs
4
1.4k
2025-12-27 Claude CodeでPRレビュー対応を効率化する@機械学習社会実装勉強会第54回
nakamasato
4
1.1k
AWS運用を効率化する!AWS Organizationsを軸にした一元管理の実践/nikkei-tech-talk-202512
nikkei_engineer_recruiting
0
170
日本の AI 開発と世界の潮流 / GenAI Development in Japan
hariby
1
500
Featured
See All Featured
The B2B funnel & how to create a winning content strategy
katarinadahlin
PRO
0
190
I Don’t Have Time: Getting Over the Fear to Launch Your Podcast
jcasabona
34
2.6k
RailsConf & Balkan Ruby 2019: The Past, Present, and Future of Rails at GitHub
eileencodes
141
34k
Mobile First: as difficult as doing things right
swwweet
225
10k
Designing Dashboards & Data Visualisations in Web Apps
destraynor
231
54k
A brief & incomplete history of UX Design for the World Wide Web: 1989–2019
jct
1
270
Building Better People: How to give real-time feedback that sticks.
wjessup
370
20k
Dealing with People You Can't Stand - Big Design 2015
cassininazir
367
27k
Future Trends and Review - Lecture 12 - Web Technologies (1019888BNR)
signer
PRO
0
3.1k
Noah Learner - AI + Me: how we built a GSC Bulk Export data pipeline
techseoconnect
PRO
0
74
Navigating Algorithm Shifts & AI Overviews - #SMXNext
aleyda
0
1k
Design and Strategy: How to Deal with People Who Don’t "Get" Design
morganepeng
132
19k
Transcript
)*(0#"4)*"84 ୈճ׆༻ฤ େᖒ༐ే "NB[PO1FSTPOBMJ[FͰ͡ΊΔϨίϝϯυαʔϏε
ࣗݾհ େᖒ༐ే σʔλΞφϦςΟΫεࣄۀຊ෦ΠϯςάϨʔγϣϯ෦ ػցֶशνʔϜΤϯδχΞ!େࡕ
ࠓ͢༰ wύʔιφϥΠζͱ wϨίϝϯυͱ wϨίϝϯυαʔϏεΛΣϒΞϓϦʹΈࠐΉྫ w"NB[PO1FSTPOBMJ[Fͱ w"NB[PO1FSTPOBMJ[FͷྲྀΕ w"NB[PO1FSTPOBMJ[Fͷྉۚ w"NB[PO1FSTPOBMJ[F4BNQMFT
εϥΠυޙͰೖख͢Δ͜ͱ͕ग़དྷ·͢ͷͰ ൃදதͷ༰ΛϝϞ͢Δඞཁ͋Γ·ͤΜɻ ࣸਅࡱӨΛ͢Δ߹ ϑϥογϡɾγϟολʔԻ͕ग़ͳ͍Α͏ʹྀ͍ͩ͘͝͞ Attention
ύʔιφϥΠζͱ
ύʔιφϥΠζͱ ύʔιφϥΠθʔγϣϯʢӳQFSTPOBMJ[BUJPOʣɺύʔιφϥΠζ ʢQFSTPOBMJ[Fʣͱ͍͏ಈࢺͷ໊ࢺܗͰ͋ΓɺʮԿ͔Λݸʑਓ͚ʹΧελϚΠζ͢Δ ͜ͱʯΛҙຯ͢Δɻ By ϑϦʔඦՊࣄయʰΟΩϖσΟΞʢWikipediaʣʱ
Ϩίϝϯυͱ
Ϩίϝϯυͱ ϢʔβͷΈߦಈʹ߹ΘͤͨΞΠςϜΛհ͢Δ͜ͱ
ϢʔβϕʔεϨίϝϯυ
ΞΠςϜϕʔεϨίϝϯυ
Ϩίϝϯυͱ ϢʔβͷΈߦಈʹ߹ΘͤͨΞΠςϜΛհ͢Δ w͓͢͢Ί͢Δ͜ͱͰϢʔβʹΞΠςϜͷൃݟΛଅ͢ wϦιʔεͷ༗ޮ׆༻ wϢʔβମݧͷ࠷దԽ ͳͲɹɹɹ
ϨίϝϯυαʔϏεΛ ΣϒΞϓϦʹΈࠐΉྫ
Ϩίϝϯυ"1*ͷೖग़ྗྫ {"userId": "13"} ϨίϝϯυAPI { "recommendedItems": [ “102", "209",
"12", “3” ] }
ϨίϝϯυαʔϏεΛΣϒΞϓϦʹΈࠐΉྫ API Gateway Lambda
"NB[PO1FSTPOBMJ[FΛ͏߹ͷྫ ϦΞϧλΠϜϨίϝϯσʔγϣϯ API Gateway Lambda Personalize
"NB[PO1FSTPOBMJ[FΛ͏߹ͷྫ όονϨίϝϯσʔγϣϯ API Gateway Lambda Lambda DynamoDB S3 Personalize
"NB[PO1FSTPOBMJ[Fͱ
"NB[PO1FSTPOBMJ[Fͱ ϑϧϚωʔδυͳϨίϝϯυαʔϏε wσʔλͷੵ͔ΒϞσϧͷֶशɺϨίϝϯυ"1*·ͰରԠ wϦΞϧλΠϜͱόονͰͷϨίϝϯυʹରԠ w"NB[PODPNͰഓΘΕ͖ٕͯͨज़Λ༻Ͱ͖Δ wػցֶशͷࣝෆཁ
"NB[PO1FSTPOBMJ[Fͱ https://aws.amazon.com/personalize/
"NB[PO1FSTPOBMJ[Fͷ༻ޠ wσʔληοτάϧʔϓιϦϡʔγϣϯ࡞ʹ༻͢ΔσʔληοτͷΈ ߹Θͤɻ6TFS *UFN 6TFSJUFNJOUFSBDUJPOͷछྨͷσʔληοτͰߏ wϨγϐͲͷΑ͏ʹσʔληοτΛॲཧ͠ɺͲΜͳΞϧΰϦζϜΛ͏͔ͷఆٛ wιϦϡʔγϣϯϨίϝϯυϞσϧɻσʔληοτ͔ΒϨγϐΛͱʹ࡞ wΩϟϯϖʔϯιϦϡʔγϣϯͷϗεςΟϯάڥɻϨίϝϯυ"1*
ରԠ͍ͯ͠ΔϨίϝϯυ Ϩγϐ ͷछྨ w64&3@1&340/"-*;"5*0/ࢦఆͨ͠Ϣʔβͷ͓͢͢ΊΞΠςϜ Λ্ҐOݸٻΊΔ w1&340/"-*;&%@3"/,*/(ࢦఆͨ͠Ϣʔβͷ͓͢͢Ί߹͍ ʹԠͨ͡ॱ൪ʹࢦఆͨ͠ΞΠςϜҰཡΛͳΒΔ w3&-"5&%@*5&.4ࢦఆͨ͠ΞΠςϜʹྨࣅ͢ΔΞΠςϜΛ্ҐOݸٻΊ Δ
"NB[PO1FSTPOBMJ[FͷྲྀΕ
جຊతͳྲྀΕ
σʔληοτάϧʔϓͷ࡞
σʔληοτάϧʔϓͷ࡞ σʔληοτ
σʔλಡΈࠐΈ
Πϕϯτͷੵ
Πϕϯτͷੵ https://docs.aws.amazon.com/personalize/latest/dg/recording-events.html
ιϦϡʔγϣϯ࡞
ιϦϡʔγϣϯόʔδϣϯ࡞ 'JOJTI͘͠$SFBUFTPMVUJPOWFSTJPOΛΫϦοΫ͢Δͱ ιϦϡʔγϣϯόʔδϣϯ͕֬ೝͳ͠Ͱ࡞͞ΕΔͷͰҙ ιϦϡʔγϣϯ࡞ޙ ιϦϡʔγϣϯը໘
Ωϟϯϖʔϯ࡞
ϦΞϧλΠϜϨίϝϯσʔγϣϯ https://docs.aws.amazon.com/personalize/latest/dg/getting-recommendations.html Ϛωδϝϯτίϯιʔϧ AWS Python SDK
όονϨίϝϯσʔγϣϯ
"NB[PO1FSTPOBMJ[Fͷྉۚ
ྉۚ σʔλͷऔΓࠐΈͱτϨʔχϯά wσʔλͷऔΓࠐΈ64%(# wτϨʔχϯά64%τϨʔχϯά࣌ؒ ྫ w݄ʹ߹ܭ(#ͷσʔλΛಡΈࠐΈ 64% wʹճιϦϡʔγϣϯόʔδϣϯΛ࡞ɻ̍ճͷτϨʔχϯά࣌ؒฏۉ࣌ؒɻ
ϲ݄ɻ 64%
ྉۚ ϦΞϧλΠϜϨίϝϯσʔγϣϯ ϲ݄͋ͨΓͷ514࣌ؒ͋ͨΓͷྉۚ w࠷ॳͷສ514࣌ؒ·Ͱ64% w࣍ͷສ514࣌ؒ·Ͱ64% wສ514࣌ؒҎ߱64% ྫ w ࣌ؒ
Ͱ֤࣌ؒ 514࣌ؒ ফඅ 64% ※ TPS: 1ඵ͋ͨΓͷτϥϯβΫγϣϯ
ྉۚ όονϨίϝϯσʔγϣϯ ϲ݄͋ͨΓͷϨίϝϯσʔγϣϯઍ݅͋ͨΓͷྉۚ w࠷ॳͷઍສ݅64% w࣍ͷԯઍສ݅64% wԯ݅Ҏ߱64% ྫ wϲ݄
ؒʹສϢʔβͷϨίϝϯσʔγϣϯΛ࡞ 64% ※ ϢʔβϕʔεͰ͋Εॲཧ͞ΕͨϢʔβɺΞΠςϜϕʔεͰ͋Εॲཧ͞ΕͨΞΠςϜʹΑΔྉۚ
Amazon Personalizeʹ͍ͭͯͬͱΓ͍ͨ
"NB[PO1FSTPOBMJ[F4BNQMFT
"NB[PO1FSTPOBMJ[F4BNQMFT https://github.com/aws-samples/amazon-personalize-samples/tree/master/diagnose
"NB[PO1FSTPOBMJ[F4BNQMFT https://github.com/aws-samples/amazon-personalize-samples/tree/master/diagnose
"NB[PO1FSTPOBMJ[F4BNQMFT https://github.com/aws-samples/amazon-personalize-samples/tree/master/diagnose
"NB[PO1FSTPOBMJ[F4BNQMFT https://github.com/aws-samples/amazon-personalize-samples/tree/master/diagnose
·ͱΊ
·ͱΊ "NB[PO1FSTPOBMJ[FϑϧϚωʔδυͳϨίϝϯυ αʔϏε wػցֶशͷࣝෆཁͰɺBNB[PODPNͰഓΘΕ͖ٕͯͨज़ ͕͑Δ wػցֶशͷ͕ࣝ͋Δͱνϡʔχϯά͕༰қʹͳΔ wϦΞϧλΠϜ"1*ͱͯ͠ɺόονͰ͑Δ
None