Building a Recommendation Engine with Spark and Apache PredictionIO - Speaker Deck

Tweet

Tweet

Slide 1

Slide 1 text

Open Source Machine Learning Server 2017 JJUG CCC Spring

Slide 2

Slide 2 text

Scala ੡ ɾ ػցֶशج൫ PredictionIOͱ SparkʹΑΔϨίϝϯυγ ες Ϝ

Slide 3

Slide 3 text

ػցֶशͷ͓೰Έ • ֶशɾϞσϧσʔλͷετϨʔδ • ػցֶशͷ෼ࢄॲཧϑϨʔϜϫʔΫ • σʔλ༧ଌͷWebαʔϏεʢAPIʣԽ ղܾͷબ୒ࢶ → ఻͍͑ͨ͜ͱ

Slide 4

Slide 4 text

ػցֶशͷ͓೰Έ • ֶशɾϞσϧσʔλͷετϨʔδ • ػցֶशͷ෼ࢄॲཧϑϨʔϜϫʔΫ • σʔλ༧ଌͷWebαʔϏεʢAPIʣԽ ղܾͷબ୒ࢶ → ఻͍͑ͨ͜ͱ

Slide 5

Slide 5 text

Warning Java ͷ࿩͸͋Γ·ͤΜ Scala ੡ɾػցֶशج൫ PredictionIO ͷ͓࿩Ͱ͢ ػցֶश ͷ͜ͱ͸࿩͠·ͤΜ ͋͘·Ͱػցֶशج൫ͷ͓࿩Ͱ͢

Slide 6

Slide 6 text

Warning Java ͷ࿩͸͋Γ·ͤΜ Scala ੡ɾػցֶशج൫ PredictionIO ͷ͓࿩Ͱ͢ ػցֶश ͷ͜ͱ͸࿩͠·ͤΜ ͋͘·Ͱػցֶशج൫ͷ͓࿩Ͱ͢

Slide 7

Slide 7 text

Warning Java ͷ࿩͸͋Γ·ͤΜ Scala ੡ɾػցֶशج൫ PredictionIO ͷ͓࿩Ͱ͢ ػցֶश ͷ͜ͱ͸࿩͠·ͤΜ ͋͘·Ͱػցֶशج൫ͷ͓࿩Ͱ͢

Slide 8

Slide 8 text

Photo by Bernard Spragg. NZ About me

Slide 9

Slide 9 text

Proﬁle Takahiro Hagino Bizreach גࣜձࣾϏζϦʔν • ٻਓݕࡧΤϯδϯʮελϯόΠʯ • AIࣨ

Slide 10

Slide 10 text

Open Source Machine Learning Server

Slide 11

Slide 11 text

ٻਓݕࡧΤϯδϯ Play on Scala ͰͷϚΠΫϩαʔϏεɾΞʔΩςΫνϟ σʔλɾετϨʔδɺ ෼ࢄݕࡧʹElasticsearch Λ࠾༻ ৗ࣌400ສ݅Ҏ্ͷٻਓΛΫϩʔϦϯά iOS/AndroidΞϓϦͰ͸஍ਤݕࡧ͕Մೳ ελϯόΠ

Slide 12

Slide 12 text

x ML ٻ৬ऀͱٻਓͷϚονϯά Search Quality and Recommendation ೥ऩਪఆ Salary Prediction ۀछɾ৬छਪఆ Job Category Prediction ٻਓಛ௃ਪఆ Prediction of Job Characteristics

Slide 13

Slide 13 text

ٻ৬ऀͱٻਓͷϚονϯά Search Quality and Recommendation ೥ऩਪఆ Salary Prediction ۀछɾ৬छਪఆ Job Category Prediction ٻਓಛ௃ਪఆ Prediction of Job Characteristics x ML

Slide 14

Slide 14 text

ٻ৬ऀͱٻਓͷϚονϯά Search Quality and Recommendation ೥ऩਪఆ Salary Prediction ۀछɾ৬छਪఆ Job Category Prediction ٻਓಛ௃ਪఆ Prediction of Job Characteristics x ML

Slide 15

Slide 15 text

ٻ৬ऀͱٻਓͷϚονϯά Search Quality and Recommendation ೥ऩਪఆ Salary Prediction ۀछɾ৬छਪఆ Job Category Prediction ٻਓಛ௃ਪఆ Prediction of Job Characteristics x ML

Slide 16

Slide 16 text

ٻ৬ऀͱٻਓͷϚονϯά Search Quality and Recommendation ೥ऩਪఆ Salary Prediction ۀछɾ৬छਪఆ Job Category Prediction ٻਓಛ௃ਪఆ Prediction of Job Characteristics x ML

Slide 17

Slide 17 text

ٻ৬ऀͱٻਓͷϚονϯά Search Quality and Recommendation ೥ऩਪఆ Salary Prediction ۀछɾ৬छਪఆ Job Category Prediction ٻਓಛ௃ਪఆ Prediction of Job Characteristics x ML

Slide 18

Slide 18 text

ٻ৬ऀͱٻਓͷϚονϯά Search Quality and Recommendation ೥ऩਪఆ Salary Prediction ۀछɾ৬छਪఆ Job Category Prediction ٻਓಛ௃ਪఆ Prediction of Job Characteristics x ML

Slide 19

Slide 19 text

x ML

Slide 20

Slide 20 text

x ML

Slide 21

Slide 21 text

τϨʔχϯάσʔλ σʔλ͝ͱES΁ͷόονΠϯϙʔτεΫϦϓτΛ४උ͍ͯͨ͠  ֶशͷ؀ڥ΍࣮ߦϑϩʔ͕୲౰ऀ͝ͱʹଐਓԽ  τϨʔχϯάσʔλͷϑΥʔϚοτ΋ఆ·͍ͬͯͳ͔ͬͨ ֶशॲཧͷ࣮ߦ࣌ؒ σʔλྔͷ૿Ճʹͱ΋ͳ͏ֶशͷ௕࣌ؒԽ x MLͷ͓ͳ΍Έ

Slide 22

Slide 22 text

τϨʔχϯάσʔλ σʔλ͝ͱES΁ͷόονΠϯϙʔτεΫϦϓτΛ४උ͍ͯͨ͠  ֶशͷ؀ڥ΍࣮ߦϑϩʔ͕୲౰ऀ͝ͱʹଐਓԽ  τϨʔχϯάσʔλͷϑΥʔϚοτ΋ఆ·͍ͬͯͳ͔ͬͨ ֶशॲཧͷ࣮ߦ࣌ؒ σʔλྔͷ૿Ճʹͱ΋ͳ͏ֶशͷ௕࣌ؒԽ x MLͷ͓ͳ΍Έ

Slide 23

Slide 23 text

ֶशϞσϧͷετϨʔδ ػೳ͝ͱʹֶशϞσϧͷอଘઌ΋ଐਓԽ ༧ଌͷWeb API ػೳ͝ͱʹTornadoͳͲͰ؆қͳAPIαʔόΛ࡞੒͓ͯ͠Γɺ  ͢͹΍͘ΞϓϦʹ൓өͰ͖ͳ͍ ༧ଌ݁ՌΛฦ͢APIͱͯ͠ESΛར༻͢ΔͳͲ͍ͯͨ͠ x MLͷ͓ͳ΍Έ

Slide 24

Slide 24 text

ֶशϞσϧͷετϨʔδ ػೳ͝ͱʹֶशϞσϧͷอଘઌ΋ଐਓԽ ༧ଌͷWeb API ػೳ͝ͱʹTornadoͳͲͰ؆қͳAPIαʔόΛ࡞੒͓ͯ͠Γɺ  ͢͹΍͘ΞϓϦʹ൓өͰ͖ͳ͍ ༧ଌ݁ՌΛฦ͢APIͱͯ͠ESΛར༻͢ΔͳͲ͍ͯͨ͠ x MLͷ͓ͳ΍Έ

Slide 25

Slide 25 text

AIࣨͷ͓ͳ΍Έ શࣾͷ༷ʑͳࣄۀͱ࿈ܞ ෳ਺ͷࣄۀ͕͋ΓɺػցֶशͷػೳΛఏڙ͍ͯ͠Δ ֤ࣄۀͱσʔλ࿈ܞͷI/FΛڞ௨Խ͍ͨ͠ ։ൃϑϩʔΛڞ௨Խ͍ͨ͠ ઐ೚ͷΠϯϑϥΤϯδχΞ΋͍ͳ͍ ఏڙ͢Δػೳ͝ͱʹɺݸผʹΠϯϑϥΛ੔͑Δͱແବ͕ଟ͍ ग़དྷΔݶΓϑϨʔϜϫʔΫԽ͍ͨ͠

Slide 26

Slide 26 text

Solution ղܾࡦΛ୳͠·ͨ͠

Slide 27

Slide 27 text

Machine Learning Stacks Apps Algorithm Processing Datastore API Server (Tornado…) Scikitlearn, SparkML … DL: Caffe2, DL4j, Tensorﬂow, Chainer … Hadoop, Spark, Storm … Elasticsearch, HBASE, Redshift …

Slide 28

Slide 28 text

Machine Learning Stacks Apps Algorithm Processing Datastore API Server Scikitlearn, SparkML … DL: Caffe2, DL4j, Tensorﬂow, Chainer … Hadoop, Spark, Storm … Elasticsearch, HBASE, Redshift … PredictionIO

Slide 29

Slide 29 text

No content

Slide 30

Slide 30 text

PredictionIO?

Slide 31

Slide 31 text

Salesforce Acquires PredictionIO

Slide 32

Slide 32 text

Salesforce Acquires PredictionIO Feb 19, 2016 - TechCrunch 4BMFTGPSDF͕ػցֶशϓϥοτϑΥʔϜͷ 1SFEJDUJPO*0Λങऩ

Slide 33

Slide 33 text

Salesforce Introduces Salesforce Einstein

Slide 34

Slide 34 text

Salesforce Acquires PredictionIO Sep 18, 2016 - TechCrunch "*ΛऔΓࠐΉ4BMFTGPSDFͷ໺๬ ػցֶशϓϥοτϑΥʔϜʮ&JOTUFJOʯΛൃද

Slide 35

Slide 35 text

The most stars repositories on Github? spark apache/spark ˒ 12.8k incubator-predictionio apache/incubator-predictionio ˒ 10.2k playframework playframework/playframework ˒ 9.3k scala scala/scala ˒ 8.2k

Slide 36

Slide 36 text

spark apache/spark ˒ 12.8k incubator-predictionio apache/incubator-predictionio ˒ 10.2k playframework playframework/playframework ˒ 9.3k scala scala/scala ˒ 8.2k The most stars repositories on Github? ˒10.2k

Slide 37

Slide 37 text

What is PredictionIO?

Slide 38

Slide 38 text

Apache PredictionIO? Apache PredictionIO (incubating) is an open source Machine Learning Server built on top of state-of-the-art open source stack for developers and data scientists create predictive engines for any machine learning task.

Slide 39

Slide 39 text

Apache PredictionIO (incubating) is an open source Machine Learning Server built on top of state-of-the-art open source stack for developers and data scientists create predictive engines for any machine learning task. Apache PredictionIO? ࠷ઌ୺ͷΦʔϓϯιʔεΛ ૊߹Θͤͨػցֶशαʔό

Slide 40

Slide 40 text

Apache PredictionIO (incubating) is an open source Machine Learning Server built on top of state-of-the-art open source stack for developers and data scientists create predictive engines for any machine learning task. Apache PredictionIO? ࠷ઌ୺ͷΦʔϓϯιʔεΛ ૊߹Θͤͨػցֶशαʔό ͲΜͳػցֶशλεΫͰ΋ ༧ଌΤϯδϯ͕ͭ͘ΕΔ

Slide 41

Slide 41 text

Apache PredictionIO let you ର৅໰୊͝ͱʹςϯϓϨʔτΛ࡞Γɺ  ͙͢ʹσϓϩΠͰ͖Δ quickly build and deploy an engine as a web service on production with customizable templates; ΫΤϦ౤͛ͯ݁ՌΛฦ͢API͕͋Δ respond to dynamic queries in real-time once deployed as a web service;

Slide 42

Slide 42 text

Apache PredictionIO let you ޡࠩͷௐ੔΍ɺධՁͷ࢓૊Έ΋͋Δ evaluate and tune multiple engine variants systematically; όον or ϦΞϧλΠϜͰ  ֶशσʔλΛొ࿥͢ΔI/F͕͋Δ unify data from multiple platforms in batch or in real-time for comprehensive predictive analytics;

Slide 43

Slide 43 text

No content

Slide 44

Slide 44 text

No content

Slide 45

Slide 45 text

No content

Slide 46

Slide 46 text

REST: EventAPI SDK: EventClient

Slide 47

Slide 47 text

No content

Slide 48

Slide 48 text

Engine Template

Slide 49

Slide 49 text

No content

Slide 50

Slide 50 text

Photo by Bernard Spragg. NZ Quick Start

Slide 51

Slide 51 text

Versions Latest Release Version v0.11.0

Slide 52

Slide 52 text

Quick Startʢ೔ຊޠʣ takezoe.hatenablog.com/entry/2017/05/11/132410

Slide 53

Slide 53 text

Installation PredictionIOͷΠϯετʔϧ ιʔε͔ΒϏϧυʢϏϧυ༻εΫϦϓτʣ ·ͨDockerΠϝʔδ΋༻ҙ͞Ε͍ͯ·͢ SparkͷΠϯετʔϧ ετϨʔδͷΠϯετʔϧ ֶशʹ࢖༻͢ΔσʔλͳͲΛ֨ೲ͢ΔͨΊͷετϨʔδ ετϨʔδ͝ͱʹอଘͰ͖Δσʔλͷछྨ͕ҟͳΓ·͢ PostgreSQL / Elasticsearch / HBase / HDFS

Slide 54

Slide 54 text

PIO CLI eventserver Launch an Event Server app Manage apps that are used by the Event Server build Build an engine at the current train Kick off a training using an engine deploy Deploy an engine as an engine server

Slide 55

Slide 55 text

eventserver app build train deploy

Slide 56

Slide 56 text

Photo by Bernard Spragg. NZ System Architecture

Slide 57

Slide 57 text

System Architecture Apache Hadoop up to 2.7.2 required only if YARN and HDFS are needed  Apache HBase up to 1.2.4 Apache Spark up to 1.6.3  for Hadoop 2.6 not Spark 2.x version Elasticsearch up to 1.7.5 not the Elasticsearch 2.x version

Slide 58

Slide 58 text

No content

Slide 59

Slide 59 text

Storage roles Meta Data Event Data Model Data ✓ ✓ ✓ ✓ ✓* ✓ ✓ LOCALFS ✓

Slide 60

Slide 60 text

HDFS

Slide 61

Slide 61 text

Scala ੡ ɾ ػցֶशج൫ PredictionIOͱ SparkʹΑΔϨίϝϯυγ ες Ϝ

Slide 62

Slide 62 text

Scala ੡ ɾ ػցֶशج൫ PredictionIOͱ SparkʹΑΔϨίϝϯυγ ες Ϝ ͔͜͜Βɺຊ୊Ͱ͢

Slide 63

Slide 63 text

Photo by Bernard Spragg. NZ Implementation of   Recommendation Engine Template

Slide 64

Slide 64 text

x ML

Slide 65

Slide 65 text

Recommendation? JOB A JOB B Cafe Waiter Shibuya JOB C View Restaurant Waiter Shibuya Startup Programmer Roppongi

Slide 66

Slide 66 text

Recommendation? JOB A JOB B Cafe Waiter Shibuya JOB C View Restaurant Waiter Shibuya Startup Programmer Roppongi

Slide 67

Slide 67 text

Recommendation? JOB A JOB B Cafe Waiter Shibuya JOB C View Restaurant Waiter Shibuya Startup Programmer Roppongi Item-Based Recommendation

Slide 68

Slide 68 text

Recommendation? ? User A User B User C

Slide 69

Slide 69 text

Recommendation? ? User A User B User C

Slide 70

Slide 70 text

Recommendation? ? User A User B User C User-based Recommendation

Slide 71

Slide 71 text

Collaborative Filtering ڠௐϑΟϧλϦϯά Ϩίϝϯυͷ୅දతͳख๏ υϝΠϯ஌͕ࣝෆཁ ར༻ऀ͕ଟ͍৔߹ʹ͸༗ར Cold-Start ໰୊

Slide 72

Slide 72 text

Collaborative Filtering Job A Job B Job C Similarity User X View Through - 1 User A View Through View 1 User B Through View Through -1 User C View View View 0.5 Recommended 1.5

Slide 73

Slide 73 text

Collaborative Filtering Job A Job B Job C Similarity User X View Through - User A View Through View 1 User B Through View Through -1 User C View View View 0.5 Recommended 1.5

Slide 74

Slide 74 text

͘Θ͘͠͸

Slide 75

Slide 75 text

System Requirements ελϯόΠͷϨίϝϯυཁ݅ ϢʔβͷΫϦοΫϩά ͓ؾʹೖΓ௥Ճϩά S3ʹϩάσʔλ্͕͕͍ͬͯΔ ֶश͸೔࣍Ͱ

Slide 76

Slide 76 text

ElasticsearchͱTasteϓϥάΠϯͰ ࡞ΔϨίϝϯυγεςϜ

Slide 77

Slide 77 text

PIOಋೖલ σʔλ༻ͷESΠϯσοΫΛຖճ࡞੒ σʔλΠϯϙʔτ༻ͷઐ༻εΫϦϓτ Elasticsearch TasteϓϥάΠϯͰ࣮ߦ ศར͕ͩ൚༻ੑɺσʔλ૿ʹΑΔ࣮ߦ͕࣌ؒ՝୊ʹ ֶश݁ՌΛόϧΫϑΝΠϧͱͯ͠ग़ྗ Elasticsearch ༧ଌͷAPIͱͯ͠ར༻ ࣮ߦϑϩʔΛγΣϧεΫϦϓτͰ؅ཧ

Slide 78

Slide 78 text

σʔλ༻ͷESΠϯσοΫΛຖճ࡞੒ σʔλΠϯϙʔτ༻ͷઐ༻εΫϦϓτ Elasticsearch TasteϓϥάΠϯͰ࣮ߦ ศར͕ͩ൚༻ੑɺσʔλ૿ʹΑΔ࣮ߦ͕࣌ؒ՝୊ʹ ֶश݁ՌΛόϧΫϑΝΠϧͱͯ͠ग़ྗ Elasticsearch ༧ଌͷAPIͱͯ͠ར༻ ࣮ߦϑϩʔΛγΣϧεΫϦϓτͰ؅ཧ PIOಋೖલ

Slide 79

Slide 79 text

Click Log Favorite Log Event Server ALS Template

Slide 80

Slide 80 text

Click Log Favorite Log Event Server ALS Template pio import

Slide 81

Slide 81 text

Click Log Favorite Log Elasticsearch v5.3 cluster Event Server ALS Template pio import Data

Slide 82

Slide 82 text

Click Log Favorite Log Elasticsearch v5.3 cluster Event Server ALS Template pio import Data Spark 2 node cluster RDD

Slide 83

Slide 83 text

Click Log Favorite Log Elasticsearch v5.3 cluster Event Server ALS Template pio import Data LOCALFS Spark 2 node cluster RDD Model

Slide 84

Slide 84 text

Click Log Favorite Log Elasticsearch v5.3 cluster Event Server ALS Template pio import Data LOCALFS Spark 2 node cluster RDD Model Query Predicted Result

Slide 85

Slide 85 text

Engine Template?

Slide 86

Slide 86 text

Engine Template

Slide 87

Slide 87 text

No content

Slide 88

Slide 88 text

D A S E D-A-S-E Data Source and Data Preparator Algorithm Serving Evaluation Metrics

Slide 89

Slide 89 text

Machine Learning Flow τϨʔχϯάσʔλ Training Data ػցֶशΞϧΰϦζϜ Machine Learning Algorithm ༧ଌϞσϧ Predictive Model લॲཧ Preprocessing Πϯϓοτσʔλ Input Data ༧ଌॲཧ Prediction Server ༧ଌ݁Ռ Predicted Result

Slide 90

Slide 90 text

Machine Learning Flow τϨʔχϯάσʔλ Training Data ػցֶशΞϧΰϦζϜ Machine Learning Algorithm ༧ଌϞσϧ Predictive Model લॲཧ Preprocessing Πϯϓοτσʔλ Input Data ༧ଌॲཧ Prediction Server ༧ଌ݁Ռ Predicted Result

Slide 91

Slide 91 text

Machine Learning Flow τϨʔχϯάσʔλ Training Data ػցֶशΞϧΰϦζϜ Machine Learning Algorithm ༧ଌϞσϧ Predictive Model લॲཧ Preprocessing Πϯϓοτσʔλ Input Data ༧ଌॲཧ Prediction Server ༧ଌ݁Ռ Predicted Result

Slide 92

Slide 92 text

Machine Learning Flow τϨʔχϯάσʔλ Training Data ػցֶशΞϧΰϦζϜ Machine Learning Algorithm ༧ଌϞσϧ Predictive Model લॲཧ Preprocessing Πϯϓοτσʔλ Input Data ༧ଌॲཧ Prediction Server ༧ଌ݁Ռ Predicted Result

Slide 93

Slide 93 text

Machine Learning Flow τϨʔχϯάσʔλ Training Data ػցֶशΞϧΰϦζϜ Machine Learning Algorithm ༧ଌϞσϧ Predictive Model લॲཧ Preprocessing Πϯϓοτσʔλ Input Data ༧ଌॲཧ Prediction Server ༧ଌ݁Ռ Predicted Result

Slide 94

Slide 94 text

Machine Learning Flow τϨʔχϯάσʔλ Training Data ػցֶशΞϧΰϦζϜ Machine Learning Algorithm ༧ଌϞσϧ Predictive Model લॲཧ Preprocessing Πϯϓοτσʔλ Input Data ༧ଌॲཧ Prediction Server ༧ଌ݁Ռ Predicted Result

Slide 95

Slide 95 text

Machine Learning Flow τϨʔχϯάσʔλ Training Data ػցֶशΞϧΰϦζϜ Machine Learning Algorithm ༧ଌϞσϧ Predictive Model લॲཧ Preprocessing Πϯϓοτσʔλ Input Data ༧ଌϞσϧ Predictive Model ༧ଌ݁Ռ Predicted Result Data Source & Preparator D

Slide 96

Slide 96 text

Machine Learning Flow τϨʔχϯάσʔλ Training Data ػցֶशΞϧΰϦζϜ Machine Learning Algorithm ༧ଌϞσϧ Predictive Model લॲཧ Preprocessing Πϯϓοτσʔλ Input Data ༧ଌϞσϧ Predictive Model ༧ଌ݁Ռ Predicted Result Data Source & Preparator D Algorithm A

Slide 97

Slide 97 text

Machine Learning Flow τϨʔχϯάσʔλ Training Data ػցֶशΞϧΰϦζϜ Machine Learning Algorithm ༧ଌϞσϧ Predictive Model લॲཧ Preprocessing Πϯϓοτσʔλ Input Data ༧ଌϞσϧ Predictive Model ༧ଌ݁Ռ Predicted Result Data Source & Preparator D Algorithm A Serving S

Slide 98

Slide 98 text

Machine Learning Flow τϨʔχϯάσʔλ Training Data ػցֶशΞϧΰϦζϜ Machine Learning Algorithm ༧ଌϞσϧ Predictive Model લॲཧ Preprocessing Πϯϓοτσʔλ Input Data ༧ଌϞσϧ Predictive Model ༧ଌ݁Ռ Predicted Result Data Source & Preparator D Algorithm A Serving S E Evaluation Metrics

Slide 99

Slide 99 text

No content

Slide 100

Slide 100 text

D

Slide 101

Slide 101 text

D A

Slide 102

Slide 102 text

No content

Slide 103

Slide 103 text

Machine Learning Flow τϨʔχϯάσʔλ Training Data ػցֶशΞϧΰϦζϜ Machine Learning Algorithm ༧ଌϞσϧ Predictive Model લॲཧ Preprocessing Πϯϓοτσʔλ Input Data ༧ଌϞσϧ Predictive Model ༧ଌ݁Ռ Predicted Result Data Source & Preparator D Algorithm A Serving S E Evaluation Metrics

Slide 104

Slide 104 text

D

Slide 105

Slide 105 text

DataSource •Event Store (Event Server) ͔ΒσʔλΛಡࠐ •TrainingDataΛฦ͢

Slide 106

Slide 106 text

No content

Slide 107

Slide 107 text

No content

Slide 108

Slide 108 text

D

Slide 109

Slide 109 text

Preparator • TrainingDataʹର͢Δલॲཧ • ಛ௃நग़ • ෳ਺AlgorithmΛར༻͢Δ৔߹ͷڞ௨ॲཧ • PreparedDataʹม׵ͯ͠Algoritmʹ౉͢

Slide 110

Slide 110 text

No content

Slide 111

Slide 111 text

No content

Slide 112

Slide 112 text

Machine Learning Flow τϨʔχϯάσʔλ Training Data ػցֶशΞϧΰϦζϜ Machine Learning Algorithm ༧ଌϞσϧ Predictive Model લॲཧ Preprocessing Πϯϓοτσʔλ Input Data ༧ଌϞσϧ Predictive Model ༧ଌ݁Ռ Predicted Result Data Source & Preparator D Algorithm A Serving S E Evaluation Metrics

Slide 113

Slide 113 text

A

Slide 114

Slide 114 text

Algorithm • train() Λ࣮૷ • ༧ଌϞσϧͷֶशΛ୲౰͢Δ • pio train ίϚϯυͰݺͼग़͞ΕΔ • HDFSʢLocalFSʣʹετΞ͞ΕΔ • predict() Λ࣮૷ • σϓϩΠޙͷΫΤϦʹରͯ͠ϦΞϧλΠϜʹݺ͹ΕΔ

Slide 115

Slide 115 text

No content

Slide 116

Slide 116 text

No content

Slide 117

Slide 117 text

No content

Slide 118

Slide 118 text

No content

Slide 119

Slide 119 text

No content

Slide 120

Slide 120 text

No content

Slide 121

Slide 121 text

Machine Learning Flow τϨʔχϯάσʔλ Training Data ػցֶशΞϧΰϦζϜ Machine Learning Algorithm ༧ଌϞσϧ Predictive Model લॲཧ Preprocessing Πϯϓοτσʔλ Input Data ༧ଌϞσϧ Predictive Model ༧ଌ݁Ռ Predicted Result Data Source & Preparator D Algorithm A Serving S E Evaluation Metrics

Slide 122

Slide 122 text

No content

Slide 123

Slide 123 text

Serving • LServeΛܧঝ • serve() Λ࣮૷

Slide 124

Slide 124 text

No content

Slide 125

Slide 125 text

Machine Learning Flow τϨʔχϯάσʔλ Training Data ػցֶशΞϧΰϦζϜ Machine Learning Algorithm ༧ଌϞσϧ Predictive Model લॲཧ Preprocessing Πϯϓοτσʔλ Input Data ༧ଌϞσϧ Predictive Model ༧ଌ݁Ռ Predicted Result Data Source & Preparator D Algorithm A Serving S E Evaluation Metrics

Slide 126

Slide 126 text

Evaluation? ࠷దͳϋΠύʔύϥϝʔλͷςετ ϋΠύʔύϥϝʔλ σʔλ͔Β͸ֶशͰ͖ͳ͍ʢਓ͕ܾؒఆ͢Δʣύϥϝʔλ ϝτϦοΫΛ༻͍ͨϋΠύʔύϥϝʔλͷௐ੔͢Δ νϡʔχϯά͸ࣗಈԽ͍ͨ͠ ΫϩεόϦσʔγϣϯ τϨʔχϯάσʔλΛ෼ׂ͠ɺͦͷҰ෦Λݕূ༻ͷσʔλͱͯ͠༻͍Δख๏ ෳ਺ճͷݕূΛߦ͏

Slide 127

Slide 127 text

Cross-validation Training Data Validation Data Training Data

Slide 128

Slide 128 text

Cross-validation Training Data Validation Data Training Data

Slide 129

Slide 129 text

Cross-validation Training Data Validation Data Training Data

Slide 130

Slide 130 text

Cross-validation Training Data x10 Validation Data Training Data

Slide 131

Slide 131 text

Grid Search Parameter B Parameter A

Slide 132

Slide 132 text

Grid Search Parameter B Parameter A

Slide 133

Slide 133 text

Grid Search Parameter B Parameter A

Slide 134

Slide 134 text

Grid Search Parameter B Parameter A

Slide 135

Slide 135 text

Grid Search Parameter B Parameter A

Slide 136

Slide 136 text

Precision@k Precision@5 / Threshold = 2.0 Predicted A B C D E

Slide 137

Slide 137 text

Precision@k Precision@5 / Threshold = 2.0 Predicted A Validation B C D E A ˒ˑˑ B ˒˒˒ X ˒˒ˑ D ˑˑˑ E ˒˒ˑ

Slide 138

Slide 138 text

Precision@k Precision@5 / Threshold = 2.0 Predicted A Validation B C D E A ˒ˑˑ B ˒˒˒ X ˒˒ˑ D ˑˑˑ E ˒˒ˑ

Slide 139

Slide 139 text

Precision@k Precision@5 / Threshold = 2.0 Predicted A Validation B C D E A ˒ˑˑ B ˒˒˒ X ˒˒ˑ D ˑˑˑ E ˒˒ˑ PositiveCount: 2.0

Slide 140

Slide 140 text

x ML 5 Jobs

Slide 141

Slide 141 text

No content

Slide 142

Slide 142 text

No content

Slide 143

Slide 143 text

Photo by Bernard Spragg. NZ Conclusion

Slide 144

Slide 144 text

τϨʔχϯάσʔλ ϦΞϧλΠϜͰ΋ɺόονͰ΋σʔλΛऔΓࠐΉI/F͕͋Δ ΞΫηετʔΫϯΛൃߦͰ͖ΔͷͰɺ֤αʔϏεͱͷ࿈ܞ͕ศར Elasticsearchͷ෼ࢄετϨʔδͷػೳΛڗडͰ͖Δ ֶशॲཧͷ࣮ߦ࣌ؒ SparkͷΫϥελΛ࢖͏ͨΊɺॲཧΛ෼ࢄֶ͠शʹ͔͔Δ࣌ؒΛ୹ॖ Open Source Machine Learning Server

Slide 145

Slide 145 text

ֶशϞσϧͷετϨʔδ ελϯόΠͰ͸LOCALFSΛར༻͍ͯ͠Δ  ϞσϧͷಛੑʹԠͯ͡HDFSΛબ୒Մೳ ༧ଌͷWeb API “pio deploy” ίϚϯυ͚ͩͰ༧ଌͷAPIΛ࡞੒Ͱ͖Δ APIαʔό͸Akka-Httpϕʔε Open Source Machine Learning Server

Slide 146

Slide 146 text

Case Studies ଞͷࣄྫ

Slide 147

Slide 147 text

ॻྨબߟ௨ա཰ - ಺ఆ཰ - ಺ఆঝ୚཰ ༧ଌ Prediction for Reject Ratio ٻਓͷ೥ऩਪఆ Salary Prediction ٻਓ಺༰ͷࣗಈੜ੒ Job description writing-bot

Slide 148

Slide 148 text

No content

Slide 149

Slide 149 text

୤ɾଐਓԽ

Slide 150

Slide 150 text

Photo by Bernard Spragg. NZ Appendix

Slide 151

Slide 151 text

ϏζϦʔνͰ͸ػցֶशΞϓϦέʔγϣϯΛ ૊৫తʹ։ൃ͍ͯ͘͠ʹ͋ͨΓɺ։ൃɾӡ༻ج൫ͱ ͯ͠"QBDIF1SFEJDUJPO*0ʹऔΓ૊ΜͰ͍·͢ɻ ೥݄ฐ͔ࣾΒ໊͕ίϛολͱͯ͠ 1SFEJDUJPO*0ͷ։ൃʹࢀՃ͢Δ͜ͱʹͳΓ·ͨ͠ɻ Shinsuke Sugaya Naoki  Takezoe Takako  Shimamoto Takahiro Hagino

Slide 152

Slide 152 text

ϏζϦʔνͰ͸ػցֶशΞϓϦέʔγϣϯΛ ૊৫తʹ։ൃ͍ͯ͘͠ʹ͋ͨΓɺ։ൃɾӡ༻ج൫ͱ ͯ͠"QBDIF1SFEJDUJPO*0ʹऔΓ૊ΜͰ͍·͢ɻ ೥݄ฐ͔ࣾΒ໊͕ίϛολͱͯ͠ 1SFEJDUJPO*0ͷ։ൃʹࢀՃ͢Δ͜ͱʹͳΓ·ͨ͠ɻ Shinsuke Sugaya Naoki  Takezoe Takako  Shimamoto Takahiro Hagino

Slide 153

Slide 153 text

How to Contribute to PIO

Slide 154

Slide 154 text

Add support for Elasticsearch 5.x

Slide 155

Slide 155 text

No content

Slide 156

Slide 156 text

jpioug.org

Slide 157

Slide 157 text

● ೔ຊ Apache PredictionIO Ϣʔβձ JPIOUG Join Us!

Slide 158

Slide 158 text

No content

Slide 159

Slide 159 text

30 6݄ FRI

Slide 160

Slide 160 text

30 6݄ FRI Open Source Machine Learning Server 02 JPIOUG Meetup 19:30 @Shibuya

Slide 161

Slide 161 text

30 6݄ FRI Open Source Machine Learning Server 02 JPIOUG Meetup 19:30 @Shibuya ʲٸืʳLT͍ͨ͠ํ

Slide 162

Slide 162 text

30 8݄ WED

Slide 163

Slide 163 text

30 8݄ WED Open Source Machine Learning Server 03 JPIOUG Meetup 19:30 @Shibuya

Slide 164

Slide 164 text

30 8݄ WED Open Source Machine Learning Server 03 JPIOUG Meetup 19:30 @Shibuya ʲΏΔืʳLT͍ͨ͠ํ

Slide 165

Slide 165 text

jpioug.org

Slide 166

Slide 166 text

Open Source Machine Learning Server 2017 JJUG CCC Spring Thank You