Scribd
Upload
19 views13 pages

15.0 Collaborative Filtering

Uploaded by

tarunsnipr47
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
19 views13 pages

15.0 Collaborative Filtering

Uploaded by

tarunsnipr47
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 13

Recommendation Systems (RS)

• Use of recommendation systems all around us.


• These systems are personalizing our web
experience, telling us:
– what to buy (Amazon),
– which movies to watch (Netflix),
– whom to be friends with (Facebook),
– which songs to listen (Spotify)
• These recommendation systems leverage our
shopping/ watching/ listening patterns and
predict what we could like in future based on our
behavior patterns.
Collaborative Filtering (CF)
• The most basic models for recommendations systems are
collaborative filtering models which are based on assumption
that people like things similar to other things they like, and
things that are liked by other people with similar taste.
• We also observed that CF is the most efficient approach in
RSs in past few years [1].
Memory-based Collaborative Filtering
• Let’s have a look, how it works?

Input:
User-item rating matrix
Tasks for Collaborative Filtering-based
Recommender System

• Which book is to be recommended to user1?


• For this, CF-based RS adopts two algorithms

User-based CF algorithm
Item-based CF algorithm
User-based CF
This method identifies users that are similar to
the queried (target) user and estimate the
desired rating to be the weighted average of
the ratings of these similar users.
Working of User-based CF
• Computation for finding k-nearest (similar)
neighbors of target user

User-User Similarity table based on Cosine Similarity


❖ User Based Collaborative filtering
approach (Contd...)

July 10, 2020


Item-based CF
Item-based collaborative filtering is one kind
of recommendation method which looks for
similar items based on the items users have
already liked or positively interacted with.
Working of Item-based CF
• Computation for finding k-nearest (similar)
neighbors of target item

User-User Similarity table based on Cosine Similarity


❖ Item Based Collaborative filtering
approach (Contd...)

July 10, 2020


Similarity Metrics

• Cosine Similarity (CS)


• Adjusted Cosine Similarity (ACS)
• Pearson Correlation (PC)
• Spearman Correlation (SC)
• Jaccard Similarity (JS)
• Euclidean Distance (ED)
• Manhattan Distance (MD)
• Mean Squared Distance (MSD)

July 10, 2020


Rating Prediction Approaches

July 10, 2020


Collaborative Filtering
Problems Rating vector PCC Cosine MSD Remarks
Few Co-rated items U=[1,2,1,1,2] 1 1 1.4 PCC and Cosine are showing
with equal ratio full similarity in spite of the
ratings. V=[2,4,2,2,4] significant difference in rating
vector.
Few Co-rated items U1=[1,2], V1=[1,2] 1 1 NAN The similarity between U1
with unequal length and V1 is more but U2 and
ratings. U2=[1,2,4,5,5,5,5,5] 0.9765 0.9971 NAN V2 look more similar.
V2=[1,2,5,5,5,5,5,5]

Few co-rated items U1=[1,1,1], V1=[3,3,3] NAN 1 1 In case of flat value rating
with flat value. vector, PCC is not computable
U2=[3,3,3], V2=[3,3,3] NAN 1 NAN and cosine return value as 1.
Only one co-rated U1=[1] NAN 1 3 PCC is not computable, Cosine
item. V1=[3] will return 1 when there is
only one co-rated item.
No co-rated item and U1=[1,0,2,0,1,0,2,0,3,0] Can’t be Can’t be Can’t be There are no co-rated items,
rating vector are not calculated calculated calculated so PCC and Cosine are not
disjoint V1=[0,1,0,2,0,1,0,2,0,3] computable.
No co-rated item and U1=[1,0,1,0,2,0,0,1,5] Can’t be Can’t be Can’t be There are no co-rated items,
rating vector are calculated calculated calculated so PCC and Cosine are not
disjoint V1=[0,3,0,4,0,4,3,0,0] computable.

You might also like