Research On Movie Recommendation System Based On Deep Learning And Multi-Objective Optimization

With the rapid advancement of the Internet and information technology,recommendation algorithms have gained widespread usage in various domains.However,traditional recommendation algorithms,like collaborative filtering,encounter challenges such as the cold start problem and sparse data.Furthermore,the recommendation target singularity problem also negatively impacts the accuracy and diversity of recommendation outcomes.Deep learning techniques,such as automatic feature learning,offer a certain degree of solutions to address the challenges of data sparsity and coldstart problems,large-scale data processing and pre-trained models.Furthermore,the use of multiobjective optimization algorithms can address the issue of homogeneous recommendation results found in traditional recommendation algorithms,thereby promoting recommendation diversity and personalization.Therefore,the paper introduces a movie recommendation algorithm that combines deep learning and multi-objective optimization techniques to recommend accurate and diverse movies for users.(1)To address the cold-start and data sparsity problems in traditional recommendation algorithms,the paper proposes MC-SVD,a hybrid recommendation algorithm based on deep learning and matrix decomposition.The MC-SVD algorithm extracts deep feature vectors from the basic attribute information of users and movies respectively using the multilayer perceptron(MLP)and convolutional neural network(CNN)deep learning models.The potential factor vectors of users and movies are obtained from user-movie rating data using Funk SVD matrix decomposition model and fused with the deep feature vectors obtained from the deep learning stage for rating prediction.The movie recommendation candidate set is constructed by the Top-k of ratings.After the experiments on Movie Lens movie dataset,it is proved that the algorithm can effectively improve the accuracy of recommendation and can alleviate the cold start and data sparsity problems.(2)To address the problem of single recommendation target in current recommendation systems,the paper proposes ACM-NNIA,a recommendation algorithm based on immune multi-objective optimization,which optimizes and improves the multi-objective optimization algorithm NNIA,introduces the method of adaptively adjusting the crossover and variation probabilities,and takes the recommendation accuracy and diversity as the objective function,and takes the movie recommendation candidate set generated by MC-SVD algorithm As the initial population,the final Top-N movie recommendation list is obtained by calculating the affinity between antibodies and taking the operations of cloning,adaptive crossover and mutation.The simulation experiments demonstrate that the algorithm is able to generate movie recommendation lists with both accuracy and diversity.Combining the above research on MC-SVD algorithm and ACM-NNIA algorithm model,as well as the requirement analysis and detailed design of the movie recommendation system based on deep learning and multi-objective optimization,the design and implementation of the movie recommendation system are finally completed.Through experiments,it is proved that the system can provide accurate and diverse movie recommendations for users.Finally,the proposed paper introduces a movie recommendation algorithm based on deep learning and multi-objective optimization,and this algorithm is implemented in the movie recommendation system.Through experiments,it is proved that the system can provide users with accurate and diverse movie recommendations.