基于感知信息素蚁群算法的电子商务消费者意图识别

发布时间：2018-06-27 21:35

  本文选题：消费者意图 + 浏览　； 参考：《五邑大学》2016年硕士论文

【摘要】：随着互联网应用在商业领域的快速普及,用户的需求体验成为互联网发展的驱动力。以电子商务系统为代表的社交网络不断发生新的变化。电子商务系统在为用户提供越来越多选择的同时,其结构也变得更加复杂,用户经常会迷失在大量的商品信息空间中,无法顺利找到自己需要的商品。电子商务推荐系统直接与用户交互,模拟商店销售人员向用户提供商品推荐服务,帮助用户找到所需商品,从而顺利完成购买过程。随着电子商务系统规模的进一步扩大,电子商务推荐系统也在飞速的发展。目前主流的电子商务推荐系统有:基于协同过滤算法的电子商务推荐系统;基于用户统计算法的电子商务推荐系统;基于知识发现算法电子商务推荐系统和基于效用的电子商务推荐系统。但这些推荐系统存在着各自的局限性,因此在电子商务推荐系统中,消费者意图的识别承担着越来越重要的作用。消费者意图的识别对于电子商务商品推荐、热点引流商品选取、网站布局以及链接的设置有至关重要的影响。目前的大部分研究都认为意图是静态的,即特定的意图是伴随着特定的环境的,因而在特定的环境下有特定的变化。然而,消费者在电子商务活动中访问和选购商品时的不确定性告诉我们消费者的意图可表现为多种形态,并且是多阶段发展的。因此,本研究用蚁群算法中的蚂蚁来表示消费者,通过蚂蚁对信息素的趋好性来模拟消费者的浏览、收藏、加入购物车和购买行为的意图。因为消费者意图为表现为商品的客观属性与消费者的主观感受的匹配,所以,我们将信息素表示为商品的客观属性和消费者感知能力的内积,其值即为表示消费者意图的信息素的浓度。我们把这种信息素称为感知信息素,用来表示消费者意图。这样就可以通过蚁群算法来呈现消费者意图的动态性和不确定性。然后,本研究通过NetLogo仿真实验以获取数据,再以神经网络来识别和验证消费者的浏览、收藏、加入购物车和购买意图。实验结果表明:在95%的显著性水平下,本研究所提出的模型将意图预测的准确度从48%提升到67%左右,可以更加准确的向用户推荐商品,具有良好的现实意义。
[Abstract]:With the rapid popularity of Internet applications in the business field, user needs experience has become the driving force of the development of the Internet. Social networks, represented by e-commerce systems, are constantly undergoing new changes. Electronic commerce system provides more and more choices for users at the same time, its structure becomes more complex, users often lose in a large number of commodity information space, can not find their own needs of goods. The E-commerce recommendation system directly interacts with the user, simulates the shop salesperson to provide the product recommendation service to the user, helps the user to find the needed product, thus completes the purchase process smoothly. With the further expansion of e-commerce system, e-commerce recommendation system is also developing rapidly. At present, the main E-commerce recommendation system includes: E-commerce recommendation system based on collaborative filtering algorithm, e-commerce recommendation system based on user statistics algorithm, and electronic commerce recommendation system based on user statistics algorithm. E-commerce recommendation system based on knowledge discovery algorithm and utility-based e-commerce recommendation system. However, these recommendation systems have their own limitations, so the recognition of consumer intention plays an increasingly important role in e-commerce recommendation systems. The identification of consumer intention is of great importance to the recommendation of e-commerce products, the selection of hot spots, the layout of websites and the setting of links. Most of the current studies suggest that the intention is static, that is, the specific intention is accompanied by the specific environment, so there is a specific change in the specific environment. However, the uncertainty when consumers visit and choose goods in e-commerce activities tells us that the intention of consumers can be expressed in a variety of forms, and is multi-stage development. Therefore, the ants in ant colony algorithm are used to represent consumers, and the intention of consumers to browse, collect, add shopping cart and purchase behavior is simulated by ants' readability to pheromone. Because consumers intend to match the objective attributes of commodities with the subjective feelings of consumers, we express pheromones as the inner product of the objective attributes of commodities and the perception ability of consumers. Its value is the concentration of pheromone that represents the intention of the consumer. We call this pheromone a perceptual pheromone, which is used to express consumer intention. In this way, ant colony algorithm can be used to show the dynamic and uncertainty of consumer intention. Then, through the NetLogo simulation experiment to obtain the data, and then the neural network to identify and verify the browsing, collecting, adding shopping cart and purchase intention. The experimental results show that the proposed model can improve the accuracy of intention prediction from 48% to 67% under the significance level of 95%. It can recommend products to users more accurately and has good practical significance.
【学位授予单位】：五邑大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：F713.55;F713.36;TP18
，

本文编号：2075354