基于复杂网络的电子商商品推荐研究.docVIP

基于复杂网络的电子商务商品推荐研究 12电商一班郭佳星 【摘 要】在信息爆炸的时代，如何找到一种准确、高效的方法，使用户能快速、准确地定位到自己所需要的信息，引起了越来越多的学者的关注。推荐系统应运而生，成为了当前解决信息过载问题的非常有潜力的方法。 本文特别选取了 B2C 电子商务作为研究和应用的对象，因为推荐系统最典型的应用就是 B2C 电子商务领域，而且具有良好的发展和应用前景。本文在分析现有的推荐算法的基础上，从 B2C 电子商务推荐系统的基本需求出发，总结了现有推荐技术对 B2C 电子商务网站的适用性，提出了基于二部图 B2C 电子商务推荐模型，并设计了该模型上的推荐算法。此模型包括在线模块和离线模块两个部分。在离线模块中对用户结点进行了聚类处理，形成了相对较少的用户类结点。用户类结点在在线模块中与原有的产品结点形成了新的二部图，然后再利用基于二部图的资源扩散算法完成推荐。 离线模块的数据处理与用户结点的聚类处理都大大的降低了在线系统的计算量，系统的效率与速度都会有较大的提高。最后利用 Amazon 和天猫某品牌店的采集数据进行实验。与协同过滤算法比较在准确率、召回率、多样性三个算法指标上，都取得了更好的效果。 【关键字】：推荐系统 B2C 电子商务 推荐算法 二部图 ABSTRACT In the era of information explosion, it has attracted more and more attention of scholars to find an accurate and efficient method, which makes users get the information they need rapidly and accurately. Recommendation systems emerge as the times require, and become a potential method to solve the problem of information overload. This paper chooses B2C e-commerce as the object of study and application, because recommendation system is typically applied in the B2C domain of electronic business affairs, and it has a good development and application prospect. Based on the analysis of existing recommendation algorithm and basic needs of B2C e-commerce, this paper propose a B2C e-commerce recommendation model based on complex network, and designs a recommendation algorithm on the model after summarizing the applicability of current recommendation technology to B2C e-commerce. This model includes online and offline modules. In the offline module, less nodes of user type is formed by clustering. In the online module, the newly formed nodes of user type form a new bipartite network with the original product data, based on which resource diffusion algorithm is realized to complete the recommendation. Data processing and the user node clustering process in the offline module would greatly reduce the computation of the online system