Research On Recommendation Algorithm Of Lightweight Service

The development of information technology has brought great convenience to people’s life. With the large increase of information in the network, it has become overloaded. In order to let users get information their need accurately, recommendation systems emerge as the times require. Because it can bring tremendous commercial value and interest, whether in academia area or industry area, recommender systems have attracted great attention. In academic area, there have appeared many efficient recommendation methods, and in the industrial area, recommender systems have been widely used in various occasions. Service recommendation is one of the application examples of recommendation systems.At present, the recommendation algorithm based on collaborative filtering is one of the widely used algorithms in recommender systems. There are two important scientific problems in the rating-based collaborative filtering recommendation process. One is the computation of the similarity of users or items; the other is the prediction value of current user to the current item. The accuracy and efficiency of existing similarity calculation methods and rating prediction methods should be improved further.This paper mainly studies the problem of recommendation based on collaborative filtering. In the calculation of similarity, prediction for unknown values and the recommendation problems of mobile ad hoc networks, this paper mainly conduct the following research work:(1) Aiming at the problems of similarity computing in the rating-based systems, this paper proposed a ratio-based similarity computating method. Comparing the scores of item co-rated by the users, we can compute the similarity between users. Through the comparison of the scores of items rated by the same users, we can compute the similarity between items. This method avoids the complex operation of current methods in computing similarity. The experimental results show that the similarity calculation method proposed in this paper is more effective than the comparision ones.(2) In order to solve the problem of prediction accuracy for unknown ratings, based on the similarity computation method proposed in previous chapter, this paper proposed a new method for the prediction of unknown value. By calculating and comparing the users’ rating values and count the numbers, this method can get the prediction values. In order to evaluate the effectiveness of the method proposed in this paper, we compared several mainly used prediciton methods. The performance of the proposed method is evaluated through a large data set of real web services. Experimental results show that our method outperforms the reference schemes in the precision of prediction value, the mean absolute error (MAE) and the computation time.(3) In order to study the service recommendation problem in mobile ad hoc networks, this paper proposes a non rating service recommendation model in the networks, and put forward a similarity calculation method for nodes, which recommend services in the networks. This paper considers that the similarity of node contains two aspects. One is the mobile terminal, and the other is the user, who contains the terminal. The terminal and the user are the objective and subjective factors of the node. According to the context information of the node, we put forward a similarity calculation method to comput the objectivity similarity of a node. At the same time, according to the information of user behavior, we put forward a calculation method to compute the subjective similarity of a node for non rating recommendation. Finally, through the experiment, this paper analysed the factors which influence the success rate of service recommendation in this networks.