Context-aware Recommendation Algorithm Based On Cloud Computing Research

With the rapid development of information technology and Internet technology,the number of people using the Internet promptly increasing,and the amount of information generated by the Internet exponentially rapidly rising,people have been into the "information explosion" and "information overload" era from the lack of information era.However,the explosive growth of information,on the one hand,brought lots of choices to the user,on the other hand also increased the difficulty of user choice that the user needs to filter out the amount of information from the irrelevant information.Based on this background,recommender system is born.And its role is to dig useful information for the user based on certain knowledge.The recommender systems,on the one hand,can help users discover items that they are interested in,or that are valuable,in order to improve the user's viewing experience,on the other hand,can show the items and services to the users interested in them through the social media platform.Therefore,recommender systems can realize a win-win situation between the user and the item provider.In recent years,relying on the rapid development of machine learning and data mining technology,the recommender systems has made great progress in theory and practice.However,with the current Internet behavior information continuously accumulated,when dealing with large data,the traditional stand-alone model emerging questions such as often faced with sparse data,poor scalability,training model time and labor and other issues and so on.It finally leads to the quality of goods or the user not high,and the algorithm needs to be improved.Specifically,the main work and contribution of this paper are as follows:(1)Combining with Spark cloud computing platform,a algorithm of matrix factorization and nearest neighbor fusion is proposed.In the process of traditional matrix factorization,after matrix factorization,the similarity information of some users and objects can be lost.By analyzing the basic principle of matrix factorization,this paper proposes the fusion of nearest neighbor technology to matrix factorization.The experiment has proved that the model can effectively improve the recommendation accuracy of traditional matrix factorization.In order to further improve the scalability of the model,combining Spark's cloud computing technology,the model can improve the computational efficiency of the model.(2)A matrix factorization recommendation algorithm for converging time context is proposed.The user's behavior and interest are changing over time.By analyzing the time-context-based and user-based collaborative filtering and object-based collaborative filtering algorithm,this paper puts forward the improved algorithm that is to extend the traditional matrix factorization recommendation model,and considering the influence of time factor on user 's interest in the extended model,come up with a matrix factorization recommendation model of fusion time context.