基于关联规则的渔业信息推荐系统设计与实现

Design and implementation of fishery information recommendation system based on association rules

为了快速便捷的获取渔业科学数据信息,基于Web日志挖掘技术对渔业科学数据共享平台用户频繁访问模式进行分析,用于发现用户访问规则,进行信息推荐服务。对分析挖掘中涉及的算法进行了分析与改进,提出了IASR(IP agent session referrer)用户识别算法和关联规则Apriori算法的改进算法,试验研究表明,IASR算法将用户识别准确性提高了13%,速度是通用算法的2倍。当事务数目大于500时,改进算法的执行效率远远优于Apriori算法,速度提高6倍以上。在此基础上,讨论了系统的关键设计与实现方法,开发了渔业信息推荐系统。系统采用JAVA、AJAX开发,数据库SQL Server2005,操作系统为Windows XP。应用结果表明,系统可使用户方便快捷地获取自己感兴趣的渔业数据信息,从而提高信息服务的质量。

In order to obtain fishery scientific data quickly and easily, this article analyzed the user’s interests in visiting fishery scientific data platforms based on data mining, mined rules and gave information recommendations according to the rules. The association rule technique, one of the commonly used algorithms for mining data analysis, attempts to find some relation of transaction items to the mass data. Based on the design of the Fisheries Information Recommendation System, the association rules mining technique was used to access the web log data of the Fishery Scientific Data Platform to find the user access pattern by data processing, pattern discovery and pattern recognition analysis procedures. Researchers analyzed and improved the algorithms involved in the mining analysis, proposed the IASR （IP Agent Session Referrer） algorithm for user identification, and introduced the technology of rewrite URL, IASR used four key informations： IP, Agent, Session and Referrer, and added session identification mechanism to recognized users in order to improve the accuracy of user identification. In the light of algorithm research on association rules, the paper proposed an improved Apriori algorithm for solving the problems of a large number of candidate set data generated in the connected computing in the process, by judging the effectiveness of the joint operation in advance to reduce the item sets joint operations, the number of the candidate item sets and iterative operations for increasing the efficiency of computation. Comparative testing was made including IASR, Apriori algorithm and improved Apriori algorithm by using the web logging data of the Fisheries Science Data Platform as the experimental object. The result of experimental research showed that IASR algorithm can improve the accuracy of user identification by 13%, and that the speed is twice as fast as that of the traditional one. The improved Apriori algorithm can greatly improve the calculation efficiency. When the transaction number is greater than 500, the improved algorithm efficiency is much better than the Apriori algorithm, improving the speed by more than 6 times. The recommendation system was designed based on the research. The overall structure of the system was designed into three layers： the service layer, business logic layer and data layer, while the service module was encapsulated in the business logic layer. Then registered users and non - registered users were provided needed information recommendation services respectively. The fishery information recommendation system was developed by using JAVA, Asynchronous JavaScript and XML, SQL Server2005 and Windows XP. The application results showed that the implementation of the system could improve the quality of information service and allow users to gain the fishery information they are interested in quickly and easily.