Prediction of customers' needs: An approach based on similarity search in transactions databases

In order to recommend products or services to customers, many organizations are very interested in knowing the needs of their customers. For that reason, recommender systems become more required in many areas, especially in e-commerce and intelligent business. In this paper, we propose to use the similarity search in transaction databases for improving the efficiency of recommender systems based on similar-like. The efficiency of the proposed system is relying on the use of metric access methods for minimizing the searching cost and for supporting the high dynamicity of transactions databases. That is, a modification of D-index method is proposed to support the search in the high dynamic environment such as transactions databases. The experiments show that the proposed method significantly reduces the search cost against the sequential search, and that the proposed method is more resistant to the databases dynamicity compared with the classical D-index.