A Novel Singularity Based Improved Tanimoto Similarity Measure for Effective Recommendation Using Collaborative Filtering
C. Selvi, E. Sivasankar · 2018
Collaborative Filtering (CF) has turned into a most prosperous approach for giving customized suggestions to the objective clients. One among the direct and productive CF approach is the memory-based approach. It finds closest neighbors to the objective client utilizing the standard similarity measures. All the standard similarity measures deal with the co-rated value of the item given by the pair of clients. So, they're not fitting to supply productive recommendations on the sparse dataset wherever the amount of co-rated items is extremely less. This paper proposes a new singularity based similarity measure, that considers all the ratings given by the client for ascertaining closest neighbors. The proposed measure proves its efficiency by analyzing with the state-of-the-art measures using benchmark MovieLens dataset. The results demonstrate that the proposed measures outperform than existing comparability measures regarding different investigation measurements like Mean Absolute Error(MAE) and Root Mean Squared Error(RMSE).