Research On Movie Recommendation Algorithm Based On Convolutional Block Attention Module-Convolutional Neural Networks Model

With the rapid growth of data in the era of big data,the phenomenon of information overload has become more and more prominent.People need to spend more time and energy to select interested movies from a large number of film resources for viewing.As one of the mainstream recommendation algorithms,collaborative filtering algorithm can effectively dealed with the phenomenon of information overload by analyzed users’ rating data to recommend movies.However,the traditional collaborative filtering algorithm has poor performance when facing the problem of data sparsity,and cannot learn the deep user and movie features,resulting in the poor recommendation quality.Therefore,based on the existing algorithm research,this dissertation proposed an improved algorithm to pre fill the data to alleviate the problem of data sparsity,and constructed a CBAM-CNN model to learn the deep feature information in the data and to improved the recommendation quality.The main contents of this dissertation are as follows:(1)Improve the prefill algorithm of Tanimoto coefficient.Aiming at the sparse problem of scoring data in data,this dissertation proposed a pre filling algorithm based on improved Tanimoto coefficient to pre fill the sparse data.First of all,according to the user’s rating and preference characteristics,calculate the difference in ratings of different users for the same movie,and judge whether the users are relevant or not based on the magnitude of the difference in ratings;Secondly,calculated their Tanimoto similarity coefficients based on user relevance to construct a user similarity matrix;Then,the singular value decomposition of the original score matrix and the user similarity are combined to calculate the score prediction value;Finally,the sparse data is filled to generate a new scoring matrix,which provides a data basis for the recommendation algorithm.The algorithm is compared with several mainstream filling algorithms on MoiveLens data set.The experimental results show that the algorithm has achieved good results.(2)Collaborative filtering algorithm based on CBAM-CNN model.Based on the completion of pre filling algorithm,this dissertation proposes a collaborative filtering algorithm based on attention mechanism and convolutional block attention module convolutional neural networks(CBAM-CNN)model to learn deep user and movie features.First of all,the multi feature attributes of users and movies are extracted and input into CBAM-CNN model respectively;Secondly,the pre filling algorithm of improved Tanimoto coefficient is used to fill the original score data,and a new score data is generated as the target value of the training model;Then,the attention mechanism is introduced to assign weights to different feature attributes of users and movies,and to strengthen important features,and add dropout layer to prevent over fitting;Finally,through continuous training and adjusting the parameters and structure of the model,the trained model is applied to the prediction and scoring task.Compared with other movie recommendation algorithms on two data sets with different sizes,the algorithm has achieved good performance in experiments.Figure[31]Table[8]Reference[60]...