Research On Srbm Rating Prediction And Recommendation Based On CNN-3C Text Classification

With the development of deep learning and personalized recommendation technologies,deep learning models such as Recurrent Neural Networks(RNN)and Convolutional Neural Network(CNN)have been applied to texts because of their simple structure and complete theory.The field of information processing is used to solve text classification problems.Restricted Boltzmann Machine(RBM)is also applied to the personalized recommendation field for scoring prediction.This paper proposes a SRBM(Similarity Restricted Boltzmann Machine)score prediction based on CNN-3C(Convolutional Neural Network-three Convolutional)text classification for the complexity of emotional data of mass review texts,data sparsity in personalized recommendation fields,and cold start issues.With the recommended method.The main research work of the dissertation is as follows:(1)To propose an improved LeNet-5 model-CNN-3C text classification model.Set up tagged corpus datasets as training sets and review datasets as test sets,and compare CNN-2C,CNN-3C and CNN-4C text classification algorithms for accuracy comparison of major text classification algorithms under different sample sizes and different sample sizes.Compared with the accuracy rate,the experimental results show that the CNN-3C algorithm is superior to other text classification algorithms.(2)Propose an improved RBM model-SRBM score prediction model,use the output of CNN-3C text classification category as the SRBM model input,train SRBM model to predict the score,and compare the accuracy of SRBM and other scoring prediction algorithms under different data sparsity conditions.Comparison between RBM and SRBM scoring algorithm under different iterations,comparison between RBM and SRBM scoring algorithm under different number of hidden units,similarity based on scores under different iterations,RBM algorithm(SRBM-score)and SRBM algorithm Comparing SRBM-score with SRBM algorithm under comparison and different number of hidden units,the experimental results show that SRBM algorithm is superior to other scoring prediction algorithms.(3)A recommendation method for SRBM score prediction based on CNN-3C text classification was proposed.Through the collected Jingdong comment data set,the relevant performance of the proposed method was analyzed under different data sparsity conditions.The experimental results showed that the proposed method was effective.Improve the recommendation accuracy rate,and at the same time,mitigate the data sparseness to some extent,expands the horizon of deep learning and personalized recommendation;...