Improved Slope One Algorithm Using Multi-weight and Auxiliary Items

Abstract

Slope one algorithm is widely used in recommendation systems. Although slope one algorithm is simple and efficient, the correlation between items and data sparsity issue are still the main issues. This paper proposes a multi-weight slope one algorithm, which obtains the correlation between items from multiple aspects, and introduces auxiliary items in relatively sparse scoring data to improve the recommendation effect. Experimental results on the MovieLens show that the recommended effect of multiple weight is better than the single weight. In the case of the sparseness of the data, the MAE value is reducing by 3% when using auxiliary items.