A novel recommender algorithm using information entropy and secondary-clustering

The accuracy of final recommender results is always affected by less active evaluating information of users. We present a secondary-clustering recommender algorithm based on information entropy in this paper. By extracting characteristic words and corresponding weights, we compute the information entropy value of each webpage text browsed by users. Moreover, we obtain the content of recommender results by twice text clustering and combine the continuous random variable of uniform distribution with the approaches of the average entropy value approximation and the logarithmic function fitting, etc. The experimental results show that the new algorithm is stable during the real system operation and improves the accuracy of final recommender results comparing to the traditional algorithms.