Research And Implementation Of Video Recommendation System Based On Mahout

Rapid development of Internet information industry,brought unexpected convenien-ce to people,and in recent years,China launched the "Internet +" concept will be applied to all areas of the Internet.Including "Internet + film",many Internet companies to set up service diversified online video sites to meet the cultural needs of more and more people.Also for this reason,the amount of video information on the network becomes very large,and users want to find the content they are interested in the mass of data has become not so easy.Recommendation technology was able to analyze a large number of user behaviour on the historical data and recommend to users interested in content,therefore,this article will applied recommend technology to the video web site,so as to effectively solve such problems.Recommendation algorithm is an important factor determines a recommendation system,content-based recommendation,collaborative filtering recommendation,rule-based recommendation are popular recommendation algorithm.In this paper,by comparing the basic thought of the three recommendation algorithm and the advantages and disadvantages of each algorithm,select collaborative filtering recommendation as the recommendation algorithm of this system.It contains two branches,user-based collaborative filtering algorithm and item-based collaborative filtering algorithm,this paper deeply studies the principle of the algorithm for the realization of the algorithm and use simulation data introduced the process of algorithm,and combined with the character-istics of video recommendation system,finally selected item-based collaborative filtering algorithm to complete the recommendation module of this system.Because of the recommendation algorithm needs to analyze vast amounts of user behaviour on the historical data,in this paper,on Hadoop platform using machine learning framework Mahout to implement the recommendation algorithm,so as to improve the efficiency of the algorithm.In addition,the new user problems is the disadvantages of collaborative filtering algorithm,which have no history behavior data users cannot be recommended.this paper proposes a new user recommendation algorithm to solve this problem,the algorithm needs to use two basic data,video scoring average top 10(Top 10 algorithm)and the result data of collaborative filtering algorithm,and using Hadoop platform to implement the Top 10 algorithm.According to the user's perspective,this system is mainly divided into two big modules,administrators and ordinary users.Using Mahout,Hadoop and Spring boot and other technology to implement the recommendation algorithm and function modules of different users.Finally,through the test,verify the feasibility and accuracy of this system,and through the proposed new user recommendation algorithm can effectively solve the problem of new users.