Research On Personalized Video Recommendation Method Based On Mixed Mode

With the rapid development of software and hardware technologies in the Internet industry,massive information has been generated on the Internet.In the face of these information,how businesses can provide effective information to users and how users can obtain the right information they need?This is the main problem in the field of information retrieval.For these problems,the traditional solution is search technology,but it depends on the information entered by the user.And users always not sure whether the input information can return the correct data or not in many cases.The emergence of recommended technologies has provided solutions for merchants and users to solve this pain point at the same time.The recommendation system adopting the recommendation technology does not require the user to input a large number of keywords,but rather analyzes the user's information and historical behavior records to give the user personalized recommendation.Recommended technology as an information retrieval technology has been widely used in society and is also a research hotspot in the current academic community.The user's actual use scenario is different,and the recommended method used is also different.Collaborative filtering recommendation method based on k-means clustering is more suitable for occasions with large user scoring data.The k-means clustering algorithm can effectively improve the speed of user or project clustering in collaborative filtering recommendation algorithm.When the amount of data is small,a good recommendation effect cannot be obtained.The method of recommending a video gene based on linear regression is suitable for a scenario where the user score data is small.But when the data is large,the recommendation effect is not good.Since the above two methods have limited application scenarios,this thesis proposes a hybrid recommendation method based on collaborative filtering and video gene.The proposed method aims to overcome the deficiencies of the two single recommendation methods and to exert the advantages of the two methods.In order to verify the effectiveness of the above method,this thesis first crawls the user data on the watercress movie.And then organizes and cleans the data and converts it into a format that the algorithm can handle.In the first step,collaborative filtering recommendation based on k-means clustering is performed.First,a user project matrix is constructed and the user similarity is analyzed.Then the clustering algorithm is used to rapidly cluster to obtain a recommendation list.In the second step,the video gene recommendation based on linear regression is performed.The video gene structure is firstly analyzed.Then combination of style preference and geographical preference is taken as the user's gene preference.And the weight value of each preference degree is determined by using a linear regression algorithm.The list of objects with high gene preference is selected to form a recommendation list.Finally,the final recommendation results are obtained by weighting the results recommended by the above two methods.Experimental results show that the hybrid recommendation method based on collaborative filtering and video gene is more accurate than the two single recommendation methods which means that the method proposed in this thesis has good generality.