Thesis Topic Details

Topic ID:
3533
Title:
Recommender Systems
Supervisor:
Helen Hye-Young Paik
Research Area:
Data Mining and Analysis
Associated Staff
Assessor:
Mike Bain
Topic Details
Status:
Active
Type:
R & D
Programs:
CS CE SE
Group Suitable:
No
Industrial:
No
Pre-requisites:
Database Systems
Description:
Recommender Systems can produce various decision-making suggestions to users, such as which movie to watch, which products to buy, which music to listen to, which destination to choose for holidays, or which news article or online book to read. Recommender Systems try to predict what the most suitable products or services are based on the users' interests and help users to deal with information overload. They create useful suggestions for users and make them come across some interesting items, which otherwise they need to spend time and effort searching for them.

Collaborative filtering algorithms commonly rely on users' interaction with the system and are used in many recommender systems. They try to automatically cluster users in groups that share the same interests and use the common behaviour of the group in order to make suggestions.

Scalability is a major issue that such algorithms need to deal with, since collaborative filtering algorithms need a lot of calculations. Another issue is called cold-start. When a new item is entered into the system, it has not been rated by any user. Therefore, collaborative filtering algorithms cannot provide recommendations for items with no or very few ratings. Moreover, when a new user enters into the system and has rated no items or has rated items that no one rated them, collaborative filtering algorithms cannot understand what the user is inter¬ested in. Therefore, they cannot recommend any items to that user.

In this project we aim at implementing effective collaborative algorithms based on different application domains such as e-commerce, digital libraries, location-based systems and similar applications. We use standard datasets in order to test the effectiveness of the implemented algorithms. The algorithms can be implemented using a programming language and tested against available datasets, or some available tools such as Weka, R, SPSS Clementine could be used.
Comments:
This project is proposed as a joint thesis project with Dr. Mehregan Mahdavi who is a visiting researcher at CSE. He will also be part of the supervision team and work closely with us on this project.
Past Student Reports
 
No Reports Available. Contact the supervisor for more information.

Check out all available reports in the CSE Thesis Report Library.

NOTE: only current CSE students can login to view and select reports to download.