On the basis of expert system, we design a costume recommendation system which provides customers with clothing collocation solution and more experience. We set up a costume matching knowledge base collected from experts, and represent the knowledge with production rules. By analyzing the customers’ specific physical information got through man-machine interface, the proposed system provides customers an intelligent costume recommendation strategy in accordance with blackboard model reasoning. Moreover, index adding algorithm is integrated into the traditional serial blackboard model in the system. Finally, we present experiments which show the search rate is improved significantly.

Introduction

In recent years, online sales of costume are more and more popular with the rapid development of the global economy and communication. At the same time, the return rate is rising continuously[1]. The possible reason is that consumers cannot try the clothes, so it is difficult for them to judge whether the costumes are suitable or not. In this circumstance, the customers keep a wait-and-see attitude and feel hard to make a purchase decision, which reduces the trust and loyalty for online costume sales eventually. At present, the major achievement of online costume recommendation is the recommendation based on establishing size and 3D simulation[2-4]. For example, a recommendation system named “right size” adopts the technology of “Rosetta stone” to recommend costume size. The model in the virtual fitting room can change the figure and simulate the customer according to the personal information. Then the customer can choose his favorite costumes and put them on the model to see the fitting effect. The disadvantage is that the system only offers the online fitting, not the specific costume recommendation. What’s more, the virtual model is not able to replace the real the real figure com .edu.cn pletely. In China, Zheng Aihua’s team applied the back propagation (BP) neural network to the costume size recommendation. She took the figure data as the input of the BP neural network and then output the corresponding size. Numerous samples were trained in the network to obtain the corresponding output. Finally, it realized the costume size recommendation through setting the trained threshold values and weights. But the system does not consider the costume style recommendation, which is a bit difficult to achieve satisfactory effect. Liao Chen’s team put forward the virtual dressing room system based on the cloud platform. They adopted the technology of big data and collaborative filtering algorithm to predict the customers’ clothing interests. This costume recommendation system is not widely used due to high requirements for hardware. In addition, the recommendation scheme of the system is similar to the popular websites such as Amazon, Tmall and Taobao. The recommend products are the same types as what the customers buy or search. These systems simply recommend costumes in accordance with the customer’s interests and purchasing behaviors, not considering the actual appearance of the customer and the experts’ experience. In this paper, we propose a costume recommendation system based on expert system, which provides customers clothing collocation solution.