Research And Application Of Recommendation Algorithm Based On Hybrid Strategy

With the rapid development of the Internet,the amount of information involved in all aspects of life is characterized by explosive growth,including e-commerce shopping,video entertainment and news reading.However,the problem is that the information is not all true and effective,and it needs to filter out useless information.Recommendation system plays such a role,and recommendation algorithm is the core of the recommendation system.Recommendation algorithm is to provide personalized and effective information recommendation for users by filtering out the uninteresting and useless parts from the massive content based on users’ behavior characteristics.There are many problems in the recommendation algorithm,including cold start data sparsity and so on,and the recommendation algorithm is subdivided into many kinds,each algorithm is widely used and has corresponding advantages.Aiming at the problems of the above recommendation algorithm and combining the advantages of each algorithm,this paper proposes a hybrid recommendation method based on user behavior division.First,the problem of cold start mainly exists among new users of the platform.Therefore,this paper divides users into new users and old users according to their behavior characteristics.In order to solve the traditional single high pin substances are recommended to new users personalized the disadvantage of low degree,this paper proposes a continuity of association rules based on user behavior,compared with the traditional form of association rules,the rules referred to in the preceding paragraph shall be replaced with the add to cart to buy or read the articles,and consequent still to buy items,and then to calculate the corresponding confidence level.The improved association rules more reflect the consistency of shopping behaviors,so as to make full use of users’ behavior information before shopping.For regular users,on the basis of the conventional collaborative filtering algorithm process,statistical analysis is made on the data attribute characteristics of users and items,and the model is established to output the probability of whether users buy items,and the probability is used as the basis of recommendation ordering.Coverage rate,accuracy rate,F1 and other indicators were used to evaluate the accuracy of recommendations,and hamming distance was used as an auxiliary measure of the personalized degree of recommendations,so as to achieve a comprehensive evaluation of the effect of recommendations.The experiment shows that compared with the traditional rule recommendation based on popularity,the association rule based on user behavior consistency proposed in this paper does make the recommendation effect better for new users,with the coverage rate increasing by 9.65% and the accuracy rate increasing by 6.42%.Meanwhile,it also highlights the individuation more.For regular users,the probability of output of the hybrid algorithm combining item-based collaborative filtering and machine learning is higher than the probability of output of the single item-based similarity sorting algorithm,with an average improvement of 3.62% in F1.