A BP Neural Network Recommendation Algorithm Based on Cloud Model

Abstract

Rating prediction is one of the key studies in the recommendation system. The traditional rating prediction algorithms only utilize user’s rating data to predict unknown ratings. In fact, however, we need to deal with the randomness and fuzziness of user ratings. At the same time, the sparsity of the data limits the performance of these algorithms. Therefore, a backpropagation neural network recommendation algorithm based on cloud model is proposed. First, the algorithm uses cloud model qualitative and quantitative transformation method to deal with the user ratings, which generates multiple cloud prediction values for users, and these values constitute the cloud layer. Then, the cloud layer joins the neural network, which can improve the accuracy of rating prediction. The missing value of rating matrix is filled in and recommendation list for the target user is generated. Finally, we perform experiments on the real-world data set, finding out that the proposed method achieves better results compared with the traditional recommendation methods in term of recall, precision, and F1.