Research On Hybrid Personalized Recommendation Method

With the development of the Internet,various industries have accumulated a large amount of data.For a company,it is very important to utilize the data and find value from it.For a customer,how to find valuable content from a large amount of data has become even more important.Personalized recommendation is an important aspect of data mining and is widely used in the industry.However,recommendation engines using a single recommendation algorithm are basically flawed,in order to improve this situation,the engine is required to have a better integration.A hybrid recommendation method based on frequent itemsets and collaborative filtering is proposed.Firstly,on the basis of frequent itemsets,the definition of supplementary score and enhancement score is given respectively.Secondly,the recommendation process of the collaborative filtering and the method based on frequent itemsets is described.Finally,the algorithm description of the mixed recommendation method is given.The experimental results show that the proposed method can effectively improve the quality of the recommendation.A hybrid recommendation algorithm based on weighted bipartite graph and logistic regression is proposed.The calculation method for the bipartite graph weight and user similarity is defined,and the recommendation list based on the bipartite graph is thus obtained.Then,the recommendation list is classified by logistic regression,and the items in the list are reordered by the classification result.The algorithm also proposes a balance factor that is used to comprehensively measure the accuracy and diversity of the recommended results.The experimental results show that the proposed method has a good recommendation effect.Firstly,data preprocessing is completed through business analysis and data cleaning.Secondly,in order to fully mine the user's potential interest points,the application proposes two hybrid recommendation algorithm models to calculate user similarity,neighbor user set and prediction score to generate a recommendation list.Finally,a set of universal personalized recommendation engine is designed and implemented.The recommendation results are generated through simple operations such as data input and hyper parameter setting,and the data results are displayed visually.