A Novel Model for Weibo Reposts Prediction by Using Generic Based Segmented BPNN

Abstract

We proposed a novel model for predict the Weibo reposts based on segmented BPNN and genetic algorithm. First of all, we studied the characteristics of Weibo reposts, and then proposed a segmented way for building BPNN by adding momentum to analyze the reposts data. Each segmented data sets will be considered as training data (inputs and outputs) to train the neural network. Secondly, genetic algorithm is utilized to generate a best solution to build the neural network. Accordingly, we include segmentation point, hidden layers, inputs, learn rate and momentum into a chromosome to find the best genes. Finally, we conduct extensive performance evaluations on the datasets of Weibo. As a result, the proposed model can improve the prediction accuracy of Weibo reposts.