Research And Application Of Recommendation Algorithm Based On LM-BP Neural Network

In recent years, the quantity of information is becoming more and more large with the development of the Internet technology. How the user capture the useful information from a number of data faster has become a hot topic in the field of network technology. Academics conducted a lot of research and practical work for such overload information and propose various forms of information personalized solutions. As an intelligent personalized information service system the Recommendation system has been widely used in the field of e-commerce, e-learning and digital libraries with the advantages of demand-driven, initiative and individuation and has been considered as the most promising personalized information technology.Collaborative filtering recommendation algorithm is one of the most successful recommendation systems. Although it has been applied widely over the years, it also has many problems like sparsity and cold start problems due to that the amount of goods is expanding while users evaluate only some of them, which leads that the accuracy of collaborative filtering recommendation algorithm still need to be improved.For the problem of data sparsity, a new collaborative filtering algorithm is designed, which choose the nearest neighbor set of users based on the size of the intersection of users’rating and use LM-BP neural network for the user-item rating prediction to increase the density of the scoring matrix. This method avoids the disadvantages of traditional dimension reduction method which would lead to lack of information and improve the accuracy of the score prediction. Thereby the quality of the collaborative filtering recommendation system is enhanced.The computing of the similarity is an important step in collaborative filtering. The traditional method is easy to exaggerate or reduce the similarity, thus affecting the quality of recommendation. In this paper, a method is proposed to calculate the similarity based on the entropy. It computes the information entropy of the score difference and adds to the formula with the degree of overlap and rating differences to improve the accuracy of similarity calculation. This method enhances the collaborative filtering recommendation algorithm effect.Finally, the real data sets called Movielens are used to test the effect of the collaborative filtering recommendation algorithm based on the LM-BP neural network proposed in this paper from these aspects like MAE, recall, F1 and precision. The effect of the score prediction based on LM-BP algorithm and the computation of similarity based on information entropy are both tested. The result shows that the collaborative filtering algorithm based on LM-BP neural network has better effects than traditional recommendation collaborative filtering algorithms.