The Artful Business of Data Mining: Distributed Schema-less Document-Based Databases - Speaker Deck

Tweet

Tweet

Slide 1

Slide 1 text

The Artful Business of Data Mining Distributed Schema-less Document-Based Databases Wednesday 24 April 13

Slide 2

Slide 2 text

David Coallier @davidcoallier Wednesday 24 April 13

Slide 3

Slide 3 text

Data Scientist At Engine Yard (.com) Wednesday 24 April 13

Slide 4

Slide 4 text

RDBMs Wednesday 24 April 13

Slide 5

Slide 5 text

Structure Restrictions Safety Wednesday 24 April 13

Slide 6

Slide 6 text

id name age address 1 2 3 4 5 6 7 ... david divad foo bar john jack jill ... 1 3 41 42 3315 4 8 ... 315 51 31 98 85 11 66 ... Wednesday 24 April 13

Slide 7

Slide 7 text

id name age address 1 2 3 4 5 6 7 ... david divad foo bar john jack jill ... 1 3 41 42 3315 4 8 ... 315 51 31 98 85 11 66 ... Wednesday 24 April 13

Slide 8

Slide 8 text

id name age address 1 2 3 4 5 6 7 ... david divad foo bar john jack jill ... 1 3 41 42 3315 4 8 ... 315 51 31 98 85 11 66 ... Wednesday 24 April 13

Slide 9

Slide 9 text

id name age address 1 2 3 4 5 6 7 ... david divad foo bar john jack jill ... 1 3 41 42 3315 4 8 ... 315 51 31 98 85 11 66 ... Wednesday 24 April 13

Slide 10

Slide 10 text

id name age address 1 2 3 4 5 6 7 ... david divad foo bar john jack jill ... 1 3 41 42 3315 4 8 ... 315 51 31 98 85 11 66 ... Wednesday 24 April 13

Slide 11

Slide 11 text

What If? Wednesday 24 April 13

Slide 12

Slide 12 text

id name age address phone 1 2 3 4 5 6 7 ... david divad foo bar john jack jill ... 26 27 42 31 17 128 21 ... IE US IE CA NZ DK IE ... 353 1 353 1 131 311 353 ... Wednesday 24 April 13

Slide 13

Slide 13 text

Before Moving on Wednesday 24 April 13

Slide 14

Slide 14 text

JSON Wednesday 24 April 13

Slide 15

Slide 15 text

What is JSON? Wednesday 24 April 13

Slide 16

Slide 16 text

{ "firstName": "David", "lastName": "Coallier", "age": 26, "address": { "streetAddress": "Mansfield House", "city": "Crosshaven", }, "phoneNumbers": [ { "type": "mobile", "number": "0863299999" } ] } Wednesday 24 April 13

Slide 17

Slide 17 text

What is HTTP? Wednesday 24 April 13

Slide 18

Slide 18 text

What is a Schema? Wednesday 24 April 13

Slide 19

Slide 19 text

Alternative Wednesday 24 April 13

Slide 20

Slide 20 text

Schema-less Wednesday 24 April 13

Slide 21

Slide 21 text

Does NOT Mean Structure-less Wednesday 24 April 13

Slide 22

Slide 22 text

Documents and K-V Buckets Wednesday 24 April 13

Slide 23

Slide 23 text

CouchDB Cluster of unreliable commodity hardware Wednesday 24 April 13

Slide 24

Slide 24 text

Replication Attachments Generated “random” ids Dictionary Revisions? JSON Objects HTTP CRUD Wednesday 24 April 13

Slide 25

Slide 25 text

Documents Wednesday 24 April 13

Slide 26

Slide 26 text

Wednesday 24 April 13

Slide 27

Slide 27 text

{ "_id": "131dafsd1vasd", "_rev": "12-fva32asdf", "firstName": "David", "lastName": "Coallier", "age": 26, "address": { "streetAddress": "Mansfield House", "city": "Crosshaven", }, "phoneNumbers": [ { "type": "mobile", "number": "0863299999" } ] } Wednesday 24 April 13

Slide 28

Slide 28 text

How do you ﬁnd Anything? Wednesday 24 April 13

Slide 29

Slide 29 text

Map/Reduce Wednesday 24 April 13

Slide 30

Slide 30 text

... Wednesday 24 April 13

Slide 31

Slide 31 text

Riak Wednesday 24 April 13

Slide 32

Slide 32 text

Dynamo Paper Wednesday 24 April 13

Slide 33

Slide 33 text

CAP Theorem Wednesday 24 April 13

Slide 34

Slide 34 text

Key-Value Buckets Wednesday 24 April 13

Slide 35

Slide 35 text

Differences? Wednesday 24 April 13

Slide 36

Slide 36 text

CouchDB Riak Storage Model append-only bitcask Access HTTP HTTP, PB Retrieval Views(M/R) M/R, Indexes, Search Versioning Eventual Consistency Vector Clocks Concurrency No Locking Client Resolution Replication master/master/slave replication, clustering Scaling In/Out Big Couch Built-in Management Futon/Fuxton Riak Control http://downloads.basho.com/papers/bitcask-intro.pdf http://guide.couchdb.org Wednesday 24 April 13

Slide 37

Slide 37 text

Map/Reduce Wednesday 24 April 13

Slide 38

Slide 38 text

Mapper: Reducer: Receives output from mappers Executed on document Wednesday 24 April 13

Slide 39

Slide 39 text

{ "_id": "...", "_rev": "...", "age": "26" } { "_id": "...", "_rev": "...", "age": "32", "heads": "3", } { "_id": "...", "_rev": "...", "age": "42" } { "_id": "...", "_rev": "...", "age": "17" } Wednesday 24 April 13

Slide 40

Slide 40 text

{ "_id": "...", "_rev": "...", "age": "26" } { "_id": "...", "_rev": "...", "age": "42" } { "_id": "...", "_rev": "...", "age": "17" } { "_id": "...", "_rev": "...", "age": "32", "heads": "3", } Wednesday 24 April 13

Slide 41

Slide 41 text

{ "age": "32", "heads": "3", } Wednesday 24 April 13

Slide 42

Slide 42 text

{ "_id": "...", "_rev": "...", "age": "26" } { "_id": "...", "_rev": "...", "age": "42" } { "_id": "...", "_rev": "...", "age": "17" } { "_id": "...", "_rev": "...", "age": "32", "heads": "3", } Map: ﬁnd-ages Wednesday 24 April 13

Slide 43

Slide 43 text

function find_ages(doc) { if (typeof(doc.age) != undefined) { emit(doc._id, doc.age); } } Map: ﬁnd-ages Wednesday 24 April 13

Slide 44

Slide 44 text

{ "_id": "...", "_rev": "...", "age": "26" } { "_id": "...", "_rev": "...", "age": "42" } { "_id": "...", "_rev": "...", "age": "17" } { "_id": "...", "_rev": "...", "age": "32", "heads": "3", } Map: ﬁnd-ages Wednesday 24 April 13

Slide 45

Slide 45 text

{ "_id": "...", "_rev": "...", "age": "26" } { "_id": "...", "_rev": "...", "age": "42" } { "_id": "...", "_rev": "...", "age": "17" } { "_id": "...", "_rev": "...", "age": "32", "heads": "3", } Map: ﬁnd-ages 26 32 42 17 Wednesday 24 April 13

Slide 46

Slide 46 text

Map: ﬁnd-ages 26 32 42 Reduce: sum 17 Wednesday 24 April 13

Slide 47

Slide 47 text

Reduce: sum function sum(values) { return sum(values); } Wednesday 24 April 13

Slide 48

Slide 48 text

Map: ﬁnd-ages 26 32 42 Reduce: sum 17 117 Wednesday 24 April 13

Slide 49

Slide 49 text

Mapper: Reducer: Receives output from mappers Executed on document Wednesday 24 April 13

Slide 50

Slide 50 text

So What? Wednesday 24 April 13

Slide 51

Slide 51 text

The Machines They Lurn. Wednesday 24 April 13

Slide 52

Slide 52 text

The Problem Wednesday 24 April 13

Slide 53

Slide 53 text

Statistics Example Wednesday 24 April 13

Slide 54

Slide 54 text

Mean, Std. Deviation Age Wednesday 24 April 13

Slide 55

Slide 55 text

µ = 1 n x i i=1 n ∑ Wednesday 24 April 13

Slide 56

Slide 56 text

σ = 1 n (x i − µ)2 i=1 n ∑ Wednesday 24 April 13

Slide 57

Slide 57 text

Mapper: Reducer: Receives output from mappers Executed on document Wednesday 24 April 13

Slide 58

Slide 58 text

Mapper: Reducer: Receive, process further. Retrieve values, pre-process Wednesday 24 April 13

Slide 59

Slide 59 text

{ "_id": "...", "_rev": "...", "age": "26" } { "_id": "...", "_rev": "...", "age": "32", "heads": "3", } { "_id": "...", "_rev": "...", "age": "42" } { "_id": "...", "_rev": "...", "age": "17" } Wednesday 24 April 13

Slide 60

Slide 60 text

[ [ 26, 676], [ 32, 1024], [ 42, 1764], [ 17, 289 ] ] Wednesday 24 April 13

Slide 61

Slide 61 text

/** * Our mapper function. */ map: function(doc) { emit(null, [doc.age, doc.age * doc.age]); } /** * Our reducer... */ reduce: function(keys, values, rereduce) { var N = 0; var summed = 0; var summedSquare = 0; for (var i in values) { N += 1; summed += values[i][0]; summedSquare += values[i][1]; } var mean = summed / N; var standard_deviation = Math.sqrt( (summedSquare / N) - (mean* mean) ) return [mean, standard_deviation] } Wednesday 24 April 13

Slide 62

Slide 62 text

/** * Our mapper function. */ map: function(doc) { emit(null, [doc.age, doc.age * doc.age]); } /** * Our reducer... */ reduce: function(keys, values, rereduce) { var N = values.length; var summed = sum(values.map(function(v) { return v[0]; })); var summedSquares = sum(values.map(function(v) { return v[1];})); var mean = summed / N; var standard_deviation = Math.sqrt( (summedSquares / N) - (mean*mean) ) return [mean, standard_deviation] } Wednesday 24 April 13

Slide 63

Slide 63 text

Naive Bayes Wednesday 24 April 13

Slide 64

Slide 64 text

Real Life Fraud Wednesday 24 April 13

Slide 65

Slide 65 text

P(x j = k | y = fraudulent) P(x j = k | y = normal) P(y) Wednesday 24 April 13

Slide 66

Slide 66 text

We need to: Sum , for each y to calculate P(x|y) x j = k Wednesday 24 April 13

Slide 67

Slide 67 text

We need: More than 1 mapper. Wednesday 24 April 13

Slide 68

Slide 68 text

We need 4 mappers Wednesday 24 April 13

Slide 69

Slide 69 text

Mapper #1: 1i P(x j = k | y = fraudulent) ∑ Wednesday 24 April 13

Slide 70

Slide 70 text

Mapper #2: 1i P(x j = k | y = normal) ∑ Wednesday 24 April 13

Slide 71

Slide 71 text

Mapper #3: 1i P(y = fraudulent) ∑ Wednesday 24 April 13

Slide 72

Slide 72 text

Mapper #4: 1i P(y = normal) ∑ Wednesday 24 April 13

Slide 73

Slide 73 text

Reducer Sums up results for parameters Wednesday 24 April 13

Slide 74

Slide 74 text

Cluster Analysis Wednesday 24 April 13

Slide 75

Slide 75 text

k-means Wednesday 24 April 13

Slide 76

Slide 76 text

Mapper: Reducer: Sum up the sums, get new centroids. Divide vectors into subgroups, Calculate d(p,q) between vectors, ﬁnd centroids, sum them up. Wednesday 24 April 13

Slide 77

Slide 77 text

More Problems Wednesday 24 April 13

Slide 78

Slide 78 text

Iterative Map/Reduce Wednesday 24 April 13

Slide 79

Slide 79 text

)BEPPQ Wednesday 24 April 13

Slide 80

Slide 80 text

Hadoop And Mr. Job Wednesday 24 April 13

Slide 81

Slide 81 text

mrjob Wednesday 24 April 13

Slide 82

Slide 82 text

Chain Mappers Wednesday 24 April 13

Slide 83

Slide 83 text

And... Reduce Wednesday 24 April 13

Slide 84

Slide 84 text

Back To Naïve Bayes Wednesday 24 April 13

Slide 85

Slide 85 text

P(x | y) ∝ P(x) P(k | x) k∈y ∏ Wednesday 24 April 13

Slide 86

Slide 86 text

class label an entry a feature P(x | y) ∝ P(x) P(k | x) k∈y ∏ Wednesday 24 April 13

Slide 87

Slide 87 text

ARGH! argmax(ln...) Wednesday 24 April 13

Slide 88

Slide 88 text

ˆ P(x) = c(N) N = N x N Wednesday 24 April 13

Slide 89

Slide 89 text

P(k | x) = c(k,x) c(k',x) k' ∑ Wednesday 24 April 13

Slide 90

Slide 90 text

ARGH! Wednesday 24 April 13

Slide 91

Slide 91 text

P(k | x) = z+c(k | x) z + c(k' | x) k' ∑ where z = { } smooth if z > 0, unsmoothed otherwise Wednesday 24 April 13

Slide 92

Slide 92 text

Wednesday 24 April 13

Slide 93

Slide 93 text

What did You see? Wednesday 24 April 13

Slide 94

Slide 94 text

c(N) N c(k, x) c(x,k') ∑ Wednesday 24 April 13

Slide 95

Slide 95 text

# This is pseudo-code def mapperPrior(self, _, line): pass def combinerPrior(self, key, values): pass def reducerPrior(self, key, values): pass def mapperProb(self, _, line): pass def combinerProb(self, key, values): pass def reducerProb(self, key, values): pass def steps(self): return [ self.mr(mapper=self.mapperPrior, combiner=self.combinerPrior, reducer=self.reducerPrior), self.mr(mapper=self.mapperProb, combiner=self.combinerProb, reducer=self.reducerProb) ] } Wednesday 24 April 13

Slide 96

Slide 96 text

Simpler Example Wednesday 24 April 13

Slide 97

Slide 97 text

from mrjob.job import MRJob class MRDoubleWordFreqCount(MRJob): def get_words(self, _, line): for word in WORD_RE.findall(line): yield word.lower(), 1 def sum_words(self, word, counts): yield word, sum(counts) def double_counts(self, word, counts): yield word, counts * 2 def steps(self): return [self.mr(mapper=self.get_words, combiner=self.sum_words, reducer=self.sum_words), self.mr(mapper=self.double_counts)] } Wednesday 24 April 13

Slide 98

Slide 98 text

Principal Component Analysis Wednesday 24 April 13

Slide 99

Slide 99 text

∑ = 1 m ( x i x i T )− µµ i=1 m ∑ T Wednesday 24 April 13

Slide 100

Slide 100 text

∑ = 1 m ( x i x i T ) i=1 m ∑ − µµT Wednesday 24 April 13

Slide 101

Slide 101 text

Summation ∑ = 1 m ( x i x i T ) i=1 m ∑ − µµT Wednesday 24 April 13

Slide 102

Slide 102 text

µ = 1 m x i i=1 m ∑ ∑ = 1 m ( x i x i T )− µµ i=1 m ∑ T Wednesday 24 April 13

Slide 103

Slide 103 text

Mappers Separate Processes Wednesday 24 April 13

Slide 104

Slide 104 text

Reducer Sum Partial Results Wednesday 24 April 13

Slide 105

Slide 105 text

Shifting Thought Paradigms Wednesday 24 April 13

Slide 106

Slide 106 text

Summation Form Wednesday 24 April 13

Slide 107

Slide 107 text

y = f(x) ∑ Wednesday 24 April 13

Slide 108

Slide 108 text

y = f(x) ∑ Mapper Reducer Wednesday 24 April 13

Slide 109

Slide 109 text

Reducer ∑ Wednesday 24 April 13

Slide 110

Slide 110 text

f(x) Mapper Wednesday 24 April 13

Slide 111

Slide 111 text

Looking Back Wednesday 24 April 13

Slide 112

Slide 112 text

Thanks Wednesday 24 April 13