Recommender Systems

25
Recommender Systems Based on: Recommender Systems by Prem Melville & Vikas Sindhwani Presented by: Vijayindu Gamage Udith Gunaratna Pubudu Gunatilaka

description

Based on: Recommender Systems by Prem Melville & Vikas Sindhwani Presented by: Vijayindu Gamage Udith Gunaratna Pubudu Gunatilaka

Transcript of Recommender Systems

Page 1: Recommender Systems

Recommender Systems

Based on:Recommender Systemsby Prem Melville & Vikas Sindhwani

Presented by:

Vijayindu Gamage

Udith Gunaratna

Pubudu Gunatilaka

Page 2: Recommender Systems

LOGORecommender Systems

Page 3: Recommender Systems

LOGORecommender Systems

Structure of a Recommender System

Page 4: Recommender Systems

LOGORecommender Systems

Classification

Collaborative Filtering Neighborhood-based Collaborative Filtering

Item-based Collaborative Filtering

Model-based Collaborative Filtering

Content Based Recommending

Hybrid Approaches

Page 5: Recommender Systems

LOGORecommender Systems

Neighborhood-based Collaborative Filtering

Basic Steps Assign a weight to all users with respect to similarity

with the active user.

Select k users that have the highest similarity with the active user – (neighborhood)

Page 6: Recommender Systems

LOGORecommender Systems

Compute a prediction from a weighted combination of

the selected neighbors’ ratings.

Page 7: Recommender Systems

LOGORecommender Systems

Neighborhood-based CF - Problem

LESS users …

neighbors are EASY to find !

Page 8: Recommender Systems

LOGORecommender Systems

Neighborhood-based CF - Problem

MANY users …

neighbors are HARD to find !

Page 9: Recommender Systems

LOGORecommender Systems

Item-based Collaborative Filtering

Proposed in 2003

DOES NOT match similar users

DOES match similar items

Leads to faster online systems

Results in improved recommendations

Page 10: Recommender Systems

LOGORecommender Systems

Item-based Collaborative Filtering

Pearson correlation is used

Rating for item i for user a is predicted

Page 11: Recommender Systems

LOGORecommender Systems

More Extensions

Highly correlated neighbors based on very few co-rated items

Significance Weighting multiply the similarity weight by a significance

weighting factor

Default Voting assume a default value for the rating for items that

have not been explicitly rated

Inverse User Frequency Universally loved/hated items are bad

Page 12: Recommender Systems

LOGORecommender Systems

Model-based Collaborative Filtering

Uses statistical models for predictions Based on data mining and machine learning

algorithms Latent factor and Matrix factorization models

have emerged as a state-of-the-art methodology

Netflix Prize competition

Page 13: Recommender Systems

LOGORecommender Systems

Content-based Recommending

Pure collaborative filtering recommenders treat all

users and items as atomic unitsCan make a better personalized recommendation

by knowing more about a user or an item Demographic information Movie genres Literary genres

Page 14: Recommender Systems

LOGORecommender Systems

Content-based Recommending

User liked

&

Movie Genre

Recommendation

Page 15: Recommender Systems

LOGORecommender Systems

Content-based Recommending

Focused on recommending items with associated textual content

2 approaches

Treat as an Information Retrieval (IR) Task

Treat as a Classification Task

Page 16: Recommender Systems

LOGORecommender Systems

Hybrid Approaches

Used to leverage the strengths of content-based and collaborative recommenders.

Merging the list results to produce a final list.

Content-boosted collaborative filtering

Page 17: Recommender Systems

LOGORecommender Systems

Evaluation Metrics

Evaluation matrix is used to measure the quality of a recommender system.

These systems are typical measured using predictive accuracy metrics

1. Mean Absolute Error (MAE)

2. Root Mean Squared Error (RMSE)

Page 18: Recommender Systems

LOGORecommender Systems

Mean Absolute Error

Where is the predicted rating for user u on item i is the actual ratingN is the total number of ratings in the test set.

Page 19: Recommender Systems

LOGORecommender Systems

Root Mean Squared Error (RMSE)

RMSE is most useful when large errors are particularly undesirable

Page 20: Recommender Systems

LOGORecommender Systems

Challenges and Limitations

SparsityCold-Start ProblemFraud

push attacks nuke attacks

Page 21: Recommender Systems

LOGORecommender Systems

Sparsity

User ratings matrix is typically very sparseEffects collaborative filtering systems

The problem

system has a very high item- to user ratio.

The system is in the initial stages of use.

Solution - making assumptions about the data generation process

Page 22: Recommender Systems

LOGORecommender Systems

Cold-Start Problem

New items and new users pose a significant challenge to recommender systems.

New item problem –

content-based approach to produce recommendations for all items,

New user problem

selecting items to be rated by a user so as to rapidly improve recommendation

performance with the least user feedback

Page 23: Recommender Systems

LOGORecommender Systems

Fraud

Push attacks Increase the rating of their own products

Nuke attacks Lower the ratings of their competitors

Item-based collaborative filtering is more robust

to these attacks Content based methods are unaffected by

profile injection attacks.

Page 24: Recommender Systems

LOGORecommender Systems

Content based or Collaborative filtering

Advantages of CF over CB

CF can perform in domains where there is not much content associated with items

CF can also preform when content is difficult for a computer to analyze.

CF system has the ability to provide serendipitous recommendations.

Page 25: Recommender Systems