Course Overview

Various aspects of recommender systems such as methods, implementation, evaluation and problems will be introduced. More specifically, representative recommendation methods such as user-based collaborative filtering, item-based collaborative filtering (association rules), content-based filtering (which represents contents of items as various numericals), knowledge-based recommendation (which requires users to show their interests) and hybrid recommendation based on machine learning using various information will be introduced. How to measure appropriateness of recommendation i.e. notion of precision, recall, novelty or serendipity for users will also be shown.

Learning Achievement

The goal of this lecture is to make students understand recommender systems and enable them to explain schema and problems of representative recommender systems.

Competence

Quantitative research competence, System expertise

Course prerequisites

Grading Philosophy

Grade of each student is determined based on the results of short test assigned in every class in manaba and term-end paper. The former consists 60% and the latter 40% of the final score. The assignments are based on course objectives and evaluated from the following viewpoints: understanding of the contents of the class, consideration, exactness of representation, and in some cases, writing structure and citation. These evaluation points are graded A, B, C and D. They are also represented as numerals and students whose total (final) scores are no less than 60 points obtain the credit of this course.

Course schedule

(1) Overview of recommender system: Methods, representative implementations and problems (Tsuji)
(2) Collaborative filtering: User-based and item-based collaborative filtering (association rule) (Tsuji)
(3) Content-based filtering: Content representation, similarity measure and use of machine learning (Tsuji)
(4) Knowledge-based recommender system: Constraint-based and case-based recommendation (Tsuji)
(5) Hybrid recommender system: Switching and combining methods of recommendation based on users' situations (Matsumura)
(6) Evaluation of recommender system: Metrics such as precision, recall, novelty, diversity and serendipity, and experimental research design (Matsumura)
(7) Various state-of-the-art recommender systems: Systems which aim to improve new evaluation metrics and which take into consideration users' situations (contexts) such as their intention and position information (Matsumura)
(8) Developing recommender system (1): Basics of programming and implementation of collaborative filtering (Matsumura)
(9) Developing recommender system (2): Implementation of user/item-based collaborative filtering (Matsumura)
(10) Developing recommender system (3): Introduction to R and its recommendation packages (Tsuji)

Course type

Lectures

Online Course Requirement

Instructor

Other information

Site for Inquiry

Link to the syllabus provided by the university