Social Network Information Recommendation Based On Distributed System

With the rapid development of Internet technology, people pay more and more attention to the information interaction of social network. In social networks, people can get a lot of important information through interaction. At the same time, with the increase in the number of social networking groups, interactive information in chatting is also growing. Lots of information constantly refresh the chat window, and the effective information rapidly covered by invalid information, people cannot quickly extract the effective information from the chat message. How to analyze the user's chat information, extract the response results, and push the answer to the user's question rapidly is the main problem solved in this paper.The business scenario of freight forwarder base on QQ logistics interaction group.According to the logistics information that the logistics company released in the QQ group, Freight Company push the appropriate logistics information to the user. Based on these business scenarios, after analyzing the demand of the information extraction from QQ group and the logistics information recommendation, this paper designs a set of social network's logistics recommendation system based on distribution system.The main work of this paper is as follows:1) Analyzing the characteristics of the QQ group logistics information, using Natural Language Processing algorithm to extract the rules of the message and obtaining the effective information to user. This paper uses HANLP as basic segmentation tool, which retooled with the pre designed word dictionaries and formative information extraction framework, to extract QQ information with user-defined rules, and ultimately get the structured logistics information. After extracting the structural logistics information, this paper also uses the geographical distance algorithm and other algorithm to perfect logistics information details to facilitate the logistics model to analyze and deal with it.2) Analyzing of the characteristics of the logistics information recommendation and designing the two layers' logistics recommendation model to recommend logistics to user. Based on distributed Mahout, the distributed recommendation model is improved by increment itemCF algorithm and the process of cold start and analysis calculation is optimized. It analyzes the historical preference of truck drivers, and recommends the appropriate logistics companies to drivers. In the second logistics model layer, according to the recommendation in the bottom layer and the quoted price information, this paper use the regression model to analyze the relationship between quoted price and logistics, and then combine this model with the logistics released by the recommended logistics companies. Finally, the two layers' logistics recommendation model generate the proper logistics that is best for drivers.3) Based on the two-layer logistics recommendation model and the HANLP pretreatment model, this paper designs a complete set of logistics information recommendation system. This system includes a QQ group's message extraction module, logistics information recommendation module and background management module. The main function of this system is parsing the QQ group message, obtaining effective logistics information and storing them into the data warehouse, then when the driver user has a new requirement,this system push the appropriate logistics to him.