Abstract
Recommendation Systems are powerful tools generating relevant suggestions for customers, as support in the decision-making process. The most sensitive step in the recommendation process is the choice of the similarity measure. The goal of this article is to present a detailed analysis of similarity measures applied to memory-based collaborative filtering techniques. Several experiments have been conducted, considering various similarity-based scenarios, to determine which measure fits best in the user-based or item-based context. Moreover, the characteristics of similarity measures and data sets (sparsity, dimensionality) are explored to determine their impact on the recommendation process. Besides, this study provides valuable information that can be used to sustain the choice of similarity measure, which can lead to improved performance of the recommendation system.
Citare
Petrusel M., A Comparative Analysis of Similarity Measures in Memory-Based Collaborative Filtering, International Conference on Artificial Intelligence and Soft Computing, 2019