Personalized Cloud Service Recommendation Based On Association Rules And User Interest Model

With the rapid development of information technologies like cloud computing, massive cloud services in the cloud platform overload.Recommendation algorithm is an effective way to solve the problem of information overload, and how to improve the computational efficiency and recommendation accuracy is an important issue in the current service recommendation research. Based on the cloud service heterogeneous information network, this paper combined association rule recommendation algorithm and user interest model to research and improve the cloud service recommendation algorithm.Firstly, the AvgSim similarity measure algorithm was improved to overcome its shortcoming of ignoring users' subjective evaluation. User ratings were considered when calculating the similarity between nodes.Based on the improved AvgSim algorithm, user interest model was established. Experiments on the MovieLens dataset are carried out and the experimental results show that the improved algorithm has promoted the accuracy of the calculation results.Secondly, the KHM clustering algorithm was studied and improved to overcome its defect of tending to converge to local optimum. The reciprocal of similarity value between two users was defined to be the square of their distance. Besides, the KHM was improved based on multi meta-paths in the cloud service heterogeneous information network.Experimental results show that the improved algorithm has achieved a better clustering result.Then, the FP-Growth association rule algorithm based on FP-Tree was studied and improved because traditional association rule algorithms lacked personalization. A new support value of nodes was defined using the weight of user's score. In addition, multi-level FP-Tree was constructed to dig the relationship in higher levels. Experimental results on the MovieLens dataset show that the improved algorithm is more accurate in recommendation result than the traditional FP-Growth algorithm.Finally, the user interest, the improved KHM clustering algorithm and the improved association rule algorithm were combined to recommend cloud services for target user. And the experimental results show that computational efficiency and recommendation accuracy are increased after the combination of these algorithms.