Research on personalized recommendation of e-commerce based on MapReduce

Research on personalized recommendation of e-commerce based on MapReduce

Bruce Lee, Beijing Jiaotong University

With the advent of the large data age, the user's personal information is distributed in various ways in different storage devices, integrating all the user information and the potential needs of users can be obtained through certain mining techniques. At present, the rapid development of E-commerce, mobile E-commerce will occupy a dominant position in the future, how to quickly tap the user's personal potential needs, users may be interested in products pushed to the user into a large data era, E-commerce enterprises need to solve the problem. At present, the accuracy of personalized recommendation of E-commerce has yet to be improved, personalized recommendation means single, and not through a deep analysis of the data, but has been based on the user's browsing information and purchase information to recommend related products to users. This recommendation method is inefficient, the personalized recommendation framework in the large data age should be analyzed and mined from the source. In view of the above problems, this article from the source of large data sets, according to the user information data sources of different ways, gather all the personality information, and then use the large data age mining technology, with the associated information mining out, and stored in the enterprise database, for enterprises to carry out related products recommended. This paper constructs the personalized recommendation system of e-commerce based on MapReduce, and gives the core module of the mining system, based on the idea of MapReduce, puts forward the method of segmenting the dataset, referred to as APD personalized recommendation algorithm, effectively avoids scanning the whole database, improves the mining efficiency, Through the theory of local optimal solution and global optimal solution, the correctness of APD personalized recommendation algorithm based on MapReduce segmentation is obtained. In the algorithm based on APD, this paper proposes a kind of gradual elimination of the idea, by setting the support degree and the confidence degree, we put forward the frequent itemsets that do not conform to the set of support degree in each independent block of data, until the optimal result is obtained, this paper gives the thought flow of the algorithm, and through the demonstration and proof of the concrete process of the algorithm, It can be concluded that this algorithm can save time and space resources, and it is of high efficiency, and is able to meet the requirements of personalized recommendation for time and space resources under large data. In the future of e-commerce recommendations, especially when mobile E-commerce occupies the initiative, the large data mining technology will be able to greatly improve the accuracy of e-commerce recommendations, will achieve good results, solve practical problems, so as to help marketers find the right marketing mix and strategy, In order to reduce the cost. Improve the success rate and profit of marketing, excavate the latent correlation and law, provide the theory basis for the scientific decision of the retailing industry in the commercial system.

Research on personalized recommendation of e-commerce based on MapReduce