Previous studies on implementing both collaborative and content based filtering systems fail to come to a conclusive solution, and in this light, the decreased accuracy of recommendations is notable. This paper shall first address methods on how to minimize the shortcomings of the two respective systems. Then, by comparing the similarity of the resulting user profiles and group profiles, it is possible to increase the accuracy of the user and group preference. To lessen the negative aspects the following must be done. With the case of the multi dimensional aspects of content based filtering, associated word mining should be used to extract relevant features. The data expressed by the mined features are not expressed as a string of data, but as a related word vector. To make up for its faults, content based filtering systems should use Bayesian classification, a system that classifies products by maintaining a knowledge base of related words. Also, to decrease the sparsity of the user-product matrix, the dimensions must be reduced. In order to reduce the dimensions of the columns, it is necessary to use Bayesian classification in tandem with the related-word knowledge base. Finally to reduce the dimensions of the rows the users must be classified into clusters.
展开▼
