A Study Of Hybrid Recommendation Algorithm Based On Collaborative Filtering And Association Rules

On account of the rapid development, extensive applications and popularization of the I nternet technology, we have stepped into the high-speed information and E-commerce era. While the internet data is growing bigger, the scales of the E-commerce enterprises based on the Internet are also expanding. The product categories and quantity are also growing rapidly. On the basis, it has produced a significant influence on the E-commerce enterprises and the individuals.In the Internet economy system, the customers are quite difficult to search for the required information, service or products at the short period of time. The users are often lost over the Internet with the unconcerned information(Resource Maze). As a result, the customers often spend a lot of time on the information searching and checking. In this way, the customers are lost. Therefore, it is much significant to conduct the recommendation algorithm based on the personalized demands in an intelligent way. The recommendation algorithm is proposed to help the Internet enterprises to solve the issues of information overload and customer loss. According to the personalized demands, the recommendation system is an advanced commercial and intelligent platform based on the mining of mass data so as to help the E-commerce enterprises to offer the customers with the personalized decision support and recommendations in the shopping process and transform the customer’s potential interests into the increases of sales and profits. However, the system has brought about more issues based on the recommendation algorithm, for instance, the coverage rate of the recommended results is relatively low with the low timeliness and precision, etc. In order to solve the above mentioned problems, this paper makes the key research on the user-based collaborative filtering algorithm and the mixed recommendation algorithm based on the strong association rule. The experiment results indicate that the recommended results are more reliable and precise based on the proposed recommendation algorithm in this paper.It is crucial to adopt the reasonable recommendation algorithms in the application of the recommendation system, which the recommendation algorithms are closely correlated with the outcomes of the system. In terms of the recommendation algorithm, this paper proposes the optimization technique of GRNN neural network on account of the user-based collaborative filtering. On the basis, it is capable of solving a series of issues that are caused by the collaborative filtering algorithm such as data sparsity, etc. At the same time, it can improve the working efficiency of the recommendation system, optimize the recommended outcomes and improve the user’s satisfaction. For this reason, the algorithm may be widely applied. In terms of the recommended algorithm based on the strong association rule, it is important to choose the threshold values of support degree and confidence coefficient. The recommended results are optimized through the selection improvements of the threshold values of support degree and confidence coefficient in the GA(genetic algorithm).In the end, the author combines the advantages of the user-based collaborative filtering algorithm and the recommendation algorithm based on the strong association rules. In order to overcome their shortages, the author proposes the mixed recommendation algorithm based on the improvements of two basic algorithms. In the proposed algorithm, the simulation data is originated from the data set of the real Tmall users. In view of the user-based collaborative filtering algorithm, the recommendation algorithm based on the strong association rule and the mixed recommendation algorithm, three experiments are proceeded to compare the corresponding indicators such as precision rates, recalling rate, MAE, time and performance, etc. The experiment results indicate that it has achieved a better performance.