Collaborative Filtering Recommendation Based on Rating Habits and Items' Attributes

Ghadah Alabduljabbar, Nouf Alrowais and Hafida Benhidour (King Saud University, Saudi Arabia)

Collaborative filtering (CF) is one of the most widely used algorithms in recommendation systems. It uses the ratings of similar users to predict the ratings of target items. Due to the sparsity of the rating matrices, the number of co-rated items becomes small which makes the computation of the similarity between users inaccurate or even impossible, thus the quality of the recommendation is affected. This paper aims at solving the problem of data sparsity in collaborative filtering. A new similarity measure that uses the users' preferences for items' attributes and the users' personal rating habits is proposed in order to improve the predicted ratings. The proposed similarity gives more weight to the users' rating habit when the number of co-rated items is high, otherwise the users' preferences for items' attributes are given a higher weight. The experimental results show that the proposed similarity has a good prediction accuracy of the ratings.

Conference: UKSim-AMSS 22nd International Conference on Computer Modelling and Simulation, UKSim2020

Published: Mar 25, 2020

DOI: 10.5013/IJSSST.a.21.02.33