Skip to Main Content
To produce high quality recommendations and achieve high coverage in the face of data sparsity in recommender systems, we explore category-based adjusted conditional probability similarity (CACPS) collaborative filtering technique in this paper. CACPS technique firstly analyzes the user-item matrix to identify relationships between different items, and then uses these relationships to indirectly compute recommendations for users. For the rating of forecasting used in recommendations, we use a weighted average in measuring the k-nearest neighbor ratings. Finally, we experimentally evaluate our results and compare them to the k-nearest neighbor approach with correlation similarity, cosine similarity and adjusted cosine similarity. Our experiments suggest that CACPS algorithms provide a better performance than the other item-based algorithms, while at the same time providing better quality than the other item-based algorithms.