Personalized recommendation algorithm based on semantic clustering

The traditional clustering algorithm doesn´t consider semantic in user session, so it cannot provide exact recommendation to users. In order to improve the recommendation quality, the thesis expresses user sessions as preferences of Web pages, then change it to semantic preferences based on domain ontology, computes semantic similarity between user sessions. Use this as a foundation, a personalized recommendation algorithm based on semantic clustering is improved, the algorithm initially clusters users with K-Nearest Neighbors of hierarchical agglomerative clustering algorithms, gets initial clustering center and the value of k, then adopts K-Means algorithm iterative refinement for clustering analysis. Experimental results show that: the improved algorithm can capture changes in user interests in time and the accuracy is better than traditional personalized recommendation algorithms.