基于关联规则的推荐系统研究-情报学专业论文.docxVIP

摘要摘 摘要 摘 要 本文针对如何实现个性化的“相关产品推荐”这～需求展开研究，引入了数 据挖掘中的关联规则技术为基本的实现手段，并在此基础上设计并实现了一套完 整的电子购物中的相关产品推荐系统ARecom(Association Rules Recommendation System)。 关联规则的算法研究部分，分为频繁项集生成和规则提取两步。针对直接决 定整体效率的频繁项集生成：本文研究了3种典型算法，并分别编程实现，比较 分析了它们在不同规模数据集上的性能特点，同时针对性能较好的FP一增长算法 在内存上的瓶颈，本文提出了一种改进的思路Fproject。关于规则提取：本文 研究了多种方法，并分析了其中负相关产生虚假规则这一弊端，针对这一问题， 提出了支持度一置信度一兴趣度相结合的框架。 在上述算法模型的基础上，设计并实现了一套相关产品推荐系统ARecom， 包括系统中Web服务器、数据库服务器、和推荐服务器的功能说明及核心技术， 最后通过～个实例证明了hRecom具有较高的效率，较准确的推荐能力。 本论文得到总装备部“交互式在线信息服务技术研究”的支持。 关键字：关联规则相关推荐频繁项集规则提取 AbstractAbstract Abstract Abstract How to realize the characteristic correlated—product recommendation is regarded as the main research target in this paper．A complete system-··--··-·ARecom of correlated-product recommendation in the field of E—conlrrlerce is designed and realized based on the association rules technology in data mining． In this paper the algorithm of association rules includes the creation of frequent itemset and the rules extraction．As to the creation of frequent itemset，this paper researches three special algorithms and realizes by programming．At the sarlle time this paper also analyzes the differentia among them in different data size and brings forward a kind of advanced idea--Fpmject as to the memory bottleneck of FP—growth algorithm．As to the rules extraction，different kinds of algorithms are brought forth． By analyzing the defect of producing misdirect-rule by negative correlation，this paper brings forward the frame that integrates support degree，confidence degree and interest degree altogether． Based on the above-mentioned algorithm model，this paper designs and realizes one complete system of correlated-product recommendation--ARecom，which includes the function and core technology of W幽Server,Data-base Sever and Recommendation Server．In the last part，the paper proves that the ARecom system has efficient and accurate recommendation abifity The paper is supposed by one of the technical basis projects of General Equipment D印artment of L．A．M w