CBPF: leveraging context and content information for better recommendations

TL;DR

This paper introduces CBPF, a context-aware recommender system that models the influence of context on ratings using Pearson Correlation, effectively integrating content and contextual data to improve recommendations.

Contribution

It presents a novel approach combining content and context with correlation modeling, evaluated across diverse datasets to demonstrate its effectiveness.

Findings

Improves recommendation accuracy with contextual information

Effective across datasets with varying sparsity levels

Modeling context influence enhances personalization

Abstract

Recommender systems help users to find their appropriate items among large volumes of information. Different types of recommender systems have been proposed. Among these, context-aware recommender systems aim at personalizing as much as possible the recommendations based on the context situation in which the user is. In this paper we present an approach integrating contextual information into the recommendation process by modeling either item-based or user-based influence of the context on ratings, using the Pearson Correlation Coefficient. The proposed solution aims at taking advantage of content and contextual information in the recommendation process. We evaluate and show effectiveness of our approach on three different contextual datasets and analyze the performances of the variants of our approach based on the characteristics of these datasets, especially the sparsity level of the input data and amount of available information.