Research On Multi-Dimensional Video Recommendation Method Based On Collaborative Filtering

With the continuous popularization of the Internet,the explosive growth of information has made it impossible for users to quickly obtain useful information when faced with a large amount of information,resulting in reduced efficiency of information usage.As one of the solutions,the recommendation system is widely used in various fields of the Internet.Most of the e-commerce platforms,video platforms,music platforms,and social platforms operate in various degrees with different recommendation strategies.The advantage of the recommendation system is that it can accurately recommend the information that the user is interested in based on the user's basic information,browsing history,interest preferences,etc.,and gradually guide the user to discover their own needs and establish a good relationship with the user.Due to the continuous complexity of the recommendation scenario and the data dimension,if the recommendation system only considers the single relationship between the user and the product,there will be some deviation between the recommendation result and the actual result.To solve the problem,it is necessary to carry out the dimension factors other than the user and the product.Analysis and selection,on the basis of the original,build a recommendation model with good predictive effect and certain practical application value.The recommendation algorithm as the core problem of the recommendation system has always been the focus of research.The traditional recommendation methods include recommendation based on collaborative filtering,recommendation based on association rules,recommendation based on decision tree,recommendation based on matrix decomposition.The recommendation method based on collaborative filtering is more commonly used,but the recommendation method only considers the user and project dimensions,ignoring implicit feedback information including user status and context,which leads to the reduction of project score prediction accuracy.Lead to a decline in the recommended effect.In order to improve the accuracy of the traditional collaborative filtering recommendation algorithm,this paper selects the LDOS-CoMoDa dataset based on the video recommendation,and combines the traditional collaborative filtering recommendation method with the multi-dimensional recommendation method to propose a collaboration-based approach.Filtered multidimensional video recommendation method.The improvement of the algorithm is as follows: Firstly,based on the traditional collaborative filtering recommendation method,the joint effect of the display feedback information and the implicit feedback information on the prediction score results in different dimensions is fully considered,and the multi-dimensional attributes are selected and determined by the regression analysis method.The degree of influence of each multi-dimensional attribute factor is linearly fitted to the multidimensional attribute factors with higher degree of influence to construct a multidimensional scoring model.Secondly,the multi-dimensional scoring model is fitted with the collaborative filtering model to construct a multi-dimensional recommendation model based on collaborative filtering.Then the prediction results of the multi-dimensional recommendation model based on collaborative filtering are compared with the actual scores of the test set data,and the method is calculated.Prediction accuracy.Finally,through the experimental method,the multi-dimensional recommendation model based on collaborative filtering proposed in this paper is compared with the collaborative filtering recommendation model and the matrixbased decomposition recommendation model.The experimental results show that the proposed model has lower MAE value than the other two traditional recommendation models.It can be considered that the video recommendation can be more accurate for users,which provides a reference for the improvement of video recommendation methods.