Realtime Recommender with Redis: Hands on

23
Realtime Recommender Hands On Torben Brodt plista GmbH June 3rd, 2013 Berlin Buzzwords http://berlinbuzzwords.de/

description

 

Transcript of Realtime Recommender with Redis: Hands on

Page 1: Realtime Recommender with Redis: Hands on

Realtime RecommenderHands On

Torben Brodtplista GmbH

June 3rd, 2013

Berlin Buzzwordshttp://berlinbuzzwords.de/

Page 2: Realtime Recommender with Redis: Hands on

Contents

1. How to feed a recommender?

2. How to pass data quickly?

3. How to build a recommender?

Page 3: Realtime Recommender with Redis: Hands on

How to feed a recommender?

Page 4: Realtime Recommender with Redis: Hands on

How to feed a recommender?

●○ recommendations & advertising network○ ~5k recs/ second

● integrated on publishers○ welt.de, abendblatt, etc

Page 5: Realtime Recommender with Redis: Hands on

How to feed a recommender?

● to show recommendations we integrated javascript● we have URL + HTTP Headers

○ user agent○ IP address -> geolocation

src http://en.wikipedia.org/wiki/Pac-Man

5

Page 6: Realtime Recommender with Redis: Hands on

How to pass data quickly?

Page 7: Realtime Recommender with Redis: Hands on

How to pass data quickly?

NOSQL!WHATELSE?

Page 8: Realtime Recommender with Redis: Hands on

How to pass data quickly?

Thrift Serialization

Page 9: Realtime Recommender with Redis: Hands on

How to pass data quickly?

Page 10: Realtime Recommender with Redis: Hands on

● Data Types: String, Lists, Set, ..● Hash

○ map between string fields and string values, very fast

○ HINCR complexity O(1)● Sorted Set

○ ZINCR complexity: O(log(N)) where N is the number of elements in the sorted set.

○ Allows to limit number of result: ZREVRANGEBYSCORE

○ UNION + INTERSECT

How to pass data quickly?p:welt.de

berlin_wins 689 +1

summer_is_coming 420

plista_company 135

10

Page 11: Realtime Recommender with Redis: Hands on

How to build a recommendation?

Page 12: Realtime Recommender with Redis: Hands on

How to build a recommendation?

Behavioralbased on interaction between user and article

○ Most Popular○ Collaborative Filtering○ Item to Item

Contentbased on the articles

○ Content Similarity○ Latest Item

Classification

● different recommender families

Page 13: Realtime Recommender with Redis: Hands on

How to build a recommendation?

welt.de/football/berlin_wins.html● ZINCR "p:welt.de" berlin_wins● ZREVRANGEBYSCORE

p:welt.de

berlin_wins 689 +1

summer_is_coming 420

plista_company 135

Live Read+ Live Write= Real Time Recommendations

Page 14: Realtime Recommender with Redis: Hands on

Most popular with timeseries

welt.de/football/berlin_wins.html● ZINCR "p:welt.de:1360007000" berlin_wins● ZUNION

○ "p:welt.de:1360007000"○ "p:welt.de:1360006000"○ "p:welt.de:1360005000"

● ZREVRANGEBYSCOREp:welt.de:1360005000

berlin_wins 420

summer_is_coming 135

plista_best_company 689

p:welt.de:1360006000

berlin_wins 420

summer_is_coming 135

plista_best_company 689

p:welt.de:1360007000

berlin_wins 689

summer_is_coming 420

plista_best_company 135

Page 15: Realtime Recommender with Redis: Hands on

Most popular with timeseries

welt.de/football/berlin_wins.html● ZINCR "p:welt.de:1360007000" berlin_wins● ZUNION ... WEIGHTS

○ "p:welt.de:1360007000" .. 4○ "p:welt.de:1360006000" .. 2○ "p:welt.de:1360005000" .. 1

● ZREVRANGEBYSCOREp:welt.de:1360005000

berlin_wins 420

summer_is_coming 135

plista_best_company 689

p:welt.de:1360006000

berlin_wins 420

summer_is_coming 135

plista_best_company 689

p:welt.de:1360007000

berlin_wins 689

summer_is_coming 420

plista_best_company 13515

Page 16: Realtime Recommender with Redis: Hands on

Most popular with timeseries

:1360007000

-1h -2h -3h -4h -5h -6h -7h -8h

:1360007000

:1360007000

42

1

Page 17: Realtime Recommender with Redis: Hands on

Most popular to any context

● it's not only publisher, we use ~50 context attributes

context attributes:● publisher● weekday● geolocation● demographics● ...

publisher = welt.de

berlin_wins 689 +1

summer_is_coming 420

plista_company 135

weekday = sunday

berlin_wins 400 +1

dortmund_wins 200

... 100

geolocation = dortmund

dortmund_wins 200

berlin_wins 10 +1

... 5

Page 18: Realtime Recommender with Redis: Hands on

Most popular to any context

ZUNION ... WEIGHTSp:welt.de:1360007 4p:welt.de:1360006 2p:welt.de:1360005 1

w:sunday:1360007 4w:sunday:1360006 2w:sunday:1360005 1

g:dortmund:1360007 4g:dortmund:1360006 2g:dortmund:1360005 1

● how it looks like in Redispublisher = welt.de

berlin_wins 689 +1

summer_is_coming 420

plista_company 135

weekday = sunday

berlin_wins 400

dortmund_wins 200

... 100

geolocation = dortmund

dortmund_wins 200

berlin_wins 10

... 5

Page 19: Realtime Recommender with Redis: Hands on

Even more Matrix Operations ;)

● Similarity Matrix

● Human Control Matrix

● Meta-learning Matrix○ cooperation with

○ aided from

∏Σ

Page 20: Realtime Recommender with Redis: Hands on

More recommenders possible

this was only about most popular

● other algorithms using redis○ incremental collaborative filtering

○ article to article paths (~graph)

○ .. using external data sources

src http://en.wikipedia.org/w

iki/Flash_(comics)

20

Page 21: Realtime Recommender with Redis: Hands on

What else in Redis?

● message bus● many recommenders● live statistics● caching

"One technology to rule them all"

Page 22: Realtime Recommender with Redis: Hands on