Affiliations: [a] Hefei Xingtai Financial Holdings (Group) Co., Ltd., Hefei, China
| [b] Post-Doctoral Station of Management Science and Engineering, University of Science and Technology of China, Hefei, China
| [c] School of Management, Guangzhou University, Guangzhou, China
| [d] Post-Doctoral Station of Applied Econometrics, Nanjing University, Nanjing, China
Correspondence:
[*]
Corresponding author. Dong Wang, School of Management, Guangzhou University, Guangzhou 510006, China. E-mail: wangdong@gzhu.edu.cn.
Abstract: Motivated by Hema Freshs new-retail case, we noticed that an effective recommender system is a common way to attract the consumers’ purchasing behaviors and thus enlarge the profit of platform as well as retailers. With the aim of increasing the benefits of all parties in the platform, this paper focusing on not only increasing the effectiveness of the recommender platform but also the evaluation system of measuring the interests of consumer, retailers and platform. In this paper, the interests of the third-party platform are added into the evaluation system, the profit of the third-party platform as an evaluation index is taken and a 0–1 integer programming model is established which sets the profit of the platform as the objective function. The result of the proposed model and algorithm indicate that: (1) The relevance of products has a significant impact on platform recommendation when the consumers are selecting products. When the correlations of the products are high, the algorithms of selecting the products will have a lower capacity of 1% compared with the algorithm without products correlations. (2) The evaluation of the target products from the target consumers is quite different from the heterogeneity assumptions. When the consumer presentation is taken into consideration, it is hard to evaluate the consumer presence because of the strictly requirement of data for the platform recommendation system. (3) The proposed two-stage solution for the platform recommendation system is optimized in time and space complexity. Total optimization of the proposed method is 30% higher than the greedy algorithms. The two stages are combined together to obtain the approximate solution, and finally provide a reasonable and feasible recommendation for the third-party platform.
