mahout recommendation with 3 columns without preferences -



mahout recommendation with 3 columns without preferences -

i have recommend videos users. have csv file containing userid, videoid, productid. under product id there many similar videos present. like:

userid videoid productid

1 2 1

1 3 1

1 5 2

2 7 2

2 8 1

2 2 1

for more clarity 1 time again factorizing :

user , video relationship:

userid videoid

1 2

1 3

1 5

2 7

2 8

2 2

consider user , video: see user 1 similar user 2 on basis of videoid 2 so, recommend user 1 watch 7 , 8 video. simple :)

but twist actual product , video info this:

videoid productid

2 1

3 1

5 2

7 2

8 1

2 1

4 1

6 1

video 4 , 6 coming under productid 1. think if user 1 come , see videoid 2 have recommend 7,8(on basis of similar user) , 4,6(on basis of similar video under same product not nowadays in actual csv).

my question is:

do need factorize csv.

what best algo it.

3.after getting result video , how rank them

what want recommend, product or video? take 1 , throw other away, don't see utilize is. recommendations come ordered , estimated preference weights.

which version of mahout recommenders utilize depends on how much info have, how many users , items. how new preference data. of mahout 0.9 recommenders can recommend users have expressed preferences , utilize preferences used calculate model.

mahout 1.0 has different mechanism can recommend anonymous or new users long have preference info them. info need not in model built mahout. method requires utilize of search engine solr or elasticsearch.

mahout docs: http://mahout.apache.org/users/recommender/intro-cooccurrence-spark.html

a preso set together: http://www.slideshare.net/pferrel/unified-recommender-39986309

mahout mahout-recommender

Comments

Popular posts from this blog

Delphi change the assembly code of a running process -

json - Hibernate and Jackson (java.lang.IllegalStateException: Cannot call sendError() after the response has been committed) -

C++ 11 "class" keyword -