A multidimensional approach for context-aware recommendation in mobile commerce

TL;DR

This paper introduces a multidimensional context-aware recommendation approach for mobile commerce that models user, item, and contextual data in a multi-space framework, improving recommendation quality.

Contribution

It presents a novel multidimensional modeling technique and a 2D recommendation space, enhancing recommendation accuracy over traditional methods.

Findings

Multidimensional approach outperforms traditional 2D methods.

Incorporating context improves recommendation relevance.

Evaluation in a restaurant system shows increased quality.

Abstract

Context as the dynamic information describing the situation of items and users and affecting the users decision process is essential to be used by recommender systems in mobile commerce to guarantee the quality of recommendation. This paper proposes a novel multidimensional approach for context aware recommendation in mobile commerce. The approach represents users, items, context information and the relationship between them in a multidimensional space. It then determines the usage patterns of each user under different contextual situations and creates a new 2 dimensional recommendation space and does the final recommendation in that space. This paper also represents an evaluation process by implementing the proposed approach in a restaurant food recommendation system considering day, time, weather and companion as the contextual information and comparing the approach with the traditional 2 dimensional one. The results of comparison illustrates that the multidimensional approach increases the recommendation quality.