Research on the Novel E-Commerce Recommendation Framework based on Modified NN Model and the Collaborative Analysis Paradigm

Ning Wang, Mingming Chen, Qiaoling Zhang, Liejun Yang
2017 International Journal of u- and e- Service, Science and Technology  
In this paper, we conduct research on novel E-commerce recommendation framework based on modified NN model and collaborative analysis. This article trusting relationship as an important attribute of influence recommendation, will use the condition transitivity of trusting relationship, designs and constructs a blending trust network, and selects the two-dimensional similar close neighbors of interest-based and trust dual factor for the target users based on this. We proposed the method based on
more » ... two essential tools, (1) while classical parametric techniques are helpless to the modeling of many natural and social phenomena, neural network techniques provide shortcuts. The function of the artificial neural network and information processing capacity is determined by structure; (2) The combination of the two algorithms can make use of the advantages of the algorithm based on content, similarity matching of the project, especially when a project that has not been without the user's evaluation that can also recommend to the user. We numerically analyze the integration model of the NN and the RS. The experiment result proves the effectiveness of the method. From the experimental simulation curve, we could conclude that the proposed method achieves better accuracy. 2 Copyright ⓒ 2017 SERSC that the recommendation system the operational mechanism can enhance the user's trust degree to the recommendation system [1] [2] [3] . Different from the traditional information filtering technology of the search engines, recommendation systems do not need to provide for search keywords, it will through the analysis of that the historical transaction records of the user or the user behavior mining potential interest, and to recommend, therefore, the recommendation system to meet the personalized requirements of users [4] [5] [6] . The following will introduce the recommended one by one interpretation. We analyze in detail in the table one.  Based on the recommendations from the basic content filtering (BCF) technology. Recommended based on content filter is the similarity between user interest and The feedforward neural network achieves structural optimization by minimizing the number of connection weights. Compared with other methods, this optimization method has several advantages: the network structure is simpler, the learning speed is faster and the induction performance is better. The structure optimization of the feedforward neural network includes the following points.
doi:10.14257/ijunesst.2017.10.1.01 fatcat:4y2xtyvu2vfgnjg55ulxflmc34