Analysis And Implementation Of Short Video Personalized Recommendation System Based On Modified Collaborative Filtering

In recent times,Science and technology are changing with each passing day,People have leapt from the era of information occlusion to the era of information explosion,how to get the data that the user is really interested in in the explosion information,That is to say,how to select the short video that users like from the trillions of short videos is a problem that needs to be solved.With the background,This paper proposes an improved collaborative video filtering short video personalized recommendation method.And this method is extended to finally realize the personalized short video recommendation system.The system makes full use of the user's registration information in the early stage,allowing the user to actively select the appropriate item(interest tag),access the third-party social account and other methods to pre-process the data,and then establish a user interest model based on the pre-processed data,and then The model matrix uses SVD to reduce the dimension,then introduces a penalty factor to calculate the similarity,and finally obtains the corresponding neighbor set,and recommends according to the prediction score.Through the introduction of SVD in the improvement process,the problem of data sparsity is solved.By introducing the penalty factor,the long tail effect caused by the project heat is effectively solved.Therefore,the improved calculation method for the traditional collaborative filtering algorithm is a more accurate and efficient method in the final recommendation result.In the process of system development,for some basic and important modules and functions,this paper gives the corresponding detailed design schemes,which provides a solid guarantee and ideas for the complete and effective implementation of the whole system.Finally,through a series of experiments,the system designed in this paper is realized and used online.At the same time,according to the actual data of the user and the relevant experimental data,the improved recommendation method is closer to the user's preference,which makes the user have a better use.Experience,which greatly increases user satisfaction and loyalty.