Design Of Video Recommendation System Based On Collaborative Filtering

The explosive growth of video resources make users face a large amount of data every day in today's Internet age,however,it is difficult to accurately obtain the data users really want only rely on the method of data search.Actually,video data overload reduces the efficiency of the use of video resources from a certain extent.It is an urgent problem to be solved to provide users with better video recommendation service through data mining.This thesis designs a video recommended system based on collaborative filtering.First,according to the collected users data,the system do data preprocessing to meet the requirements of data mining for the data source.And then,use the improved classification and regression tree to establish the user's interest model by preprocessed data source to dig out the user's personal preferences.Finally,combine two models of Item-based and User-based,the reasonable combination of the two models makes them achieve the efficient and accurate recommendation of the system on the premise of foster strengths and circumvent weaknesses.In the above process,this thesis also improves the similarity calculation formula in the traditional cooperative filtering algorithm,which effectively alleviates the influence of the popular video on the recommendation result in the recommendation process.According to the demand characteristics of the system,this thesis focuses on the function of the system in detail,which provides a clear idea for the whole system development.Experiments show that the video recommended system based on collaborative filtering is designed to provide users with more accurate recommended service by using the various improvements mentioned above.And the system can be run in the current mainstream operating system and browser through compatibility testing,which improves the user experience satisfaction.