Ch 03: U n s u p e r v i s e d L e a r n i n g - S R

Prof. Dr. Rashid A. Saeed

2
Course Content

4
Significant industrial & research interest in recommendation
systems:
2 out of 3 watched hours come from
recommendations

increases its watch times by 50% per year

35% of all sales are generated by recommendations

Netflix

6
1. Introduction

Dramatic increase of content-delivery services and multimedia

content
Wide-range of choices and tight time constraints

Efficient mechanisms required for discovering preferred content

F Improvement of user satisfaction, decrease in churn rates

Recommendation systems (RS) address these problems

ü Exploit information about user preferences
ü Accurately & efficiently recommend the appropriate content

User feedback

Explicit feedback: Users indicate their preference for items

• Ratings scales (1-5 stars, 1-10 stars)
• Binary values (liked/disliked)

Implicit feedback: Users interact with items

• Purchase history
• Viewing duration
• Click patterns

8
Ratings matrix

Objects (e.g., movies, products)

User
s

Sparse matrix: Most values are unknown

Predicting: The task of filling the unknown values

Recommendation Systems

Infer user preferences about items

Produce highly relevant & personalized lists of items

items

rating: level of
preference
users unknown
preferences
ratings matrix

10
 11

Main Approaches

Collaborative filtering (CF)

Inspect rating patters to find similar users/items

Content-based (CB)
Analyze attributes of items for building user profiles

In general, CF performs better than CB

• CF fail to provide accurate predictions with insufficient
ratings
• CB can alleviate the sparsity problem

12
Collaborative filtering (CF)

Popular Techniques
§Nearest Neighbor

§Matrix Factorization

§Deep Learning

13

Popular Recommenders: K-Νearest Νeighbors Approach

User-based collaborative filtering (UBCF)

Assumption: Similar users have similar preferences
• User similarity: agreement on co-rated items
• Prediction: weighted sum of similar user’s ratings

Item-based collaborative filtering (IBCF)

Assumption: Users have similar tastes for similar items
• Item similarity: agreement within users rated both
items
• Prediction: weighted sum of similar items’ ratings

14
14
Nearest Neighbor Using predefined Similarity Measure

Current User Users

0
1 1st item rate
Dislike 1?
0 abcd
abcd Rating
Unknown
Prediction
1 Like abcd Users
Other
1 § This user did
§ The
not prediction
rate the § There are other
1

Items
? was made
item. We will users who
Unknown 0 based
try to on
predict
the a rated the same
nearest
rating item. We are
1 neighbor.
according to interested in
his neighbors. the Nearest
1 Neighbors.

0
User ModelabcdNeighbors
Nearest
1
= interaction
We are looking
§
for the Nearest
1 abcd

1
historyNeighbor. The
one with the 1
§ Nearest
Neighbor
lowest
Hamming 0 14th item rate
distance.
Hamming 5 6 6 5 4 8
distance

15

Collaborative filtering (CF)

Popular Techniques
§Nearest Neighbor

§Matrix Factorization

§Deep Learning

16
Matrix Factorization

17

Netflix movie rating

18
Recommender Systems

19

Matrix Factorization ‫اﻟﺗﺣﻠﯾل إﻟﻰ ﻋواﻣل‬

Factorization
20=4x5

20
Categories

21

22
Matrix Factorization

23

Matrix Factorization

24
Matrix Factorization

25

Storage

26
Storage

27

Storage

28
Storage

29

30
 31

Storage

32
33

34
 System will
recommend
M5 to Dina

35

Datasets

q https://grouplens.org/datasets/movielen
s/
q https://cseweb.ucsd.edu/~jmcauley/data
sets.html
q https://www.kaggle.com/datasets/thede
vastator/book-recommender-system-
itembased
q https://cseweb.ucsd.edu/~jmcauley/data
sets.html
36
Movielens Movie Recommendation System

37

ratings

38
users

39

40
ML Model training for Recommending movies
based on users ratings

41

42