A multi-objective service composition recommendation method for individualized customer: Hybrid MPA-GSO-DNN model

作     者：Liu, Zhengchao Guo, Shunsheng Wang, Lei Du, Baigang Pang, Shibao 

作者机构：Wuhan Univ Technol Sch Mech & Elect Engn Wuhan 430070 Hubei Peoples R China 

出 版 物：《COMPUTERS & INDUSTRIAL ENGINEERING》 (计算机与工业工程)

年 卷 期：2019年第128卷

页      面：122-134页

核心收录：

学科分类：1201[管理学-管理科学与工程(可授管理学、工学学位)] 08[工学] 0812[工学-计算机科学与技术（可授工学、理学学位）] 

基　　金：National Natural Science Foundation of China [51705385, 51705386] Hubei Province Natural Science Foundation of China [2014BAA032, 2015BAA063, 2017CFB297] 

主　　题：Personalized recommendation Multi-objective preference analysis Service composition Deep learning Glowworm swarm optimization algorithm 

摘      要：With the development of manufacturing customization, unified manufacturing service recommendation is difficult to meet the customer s individualized demand. To this end, the existing research hotspots focus on solving personalized service recommendation issues. However, the personalized recommendation for service composition is more complex compared with the existing single service recommendation. Especially in the case of less customer historical data, analyzing customer preference and recommending appropriate composite service is a difficult problem. Therefore, this paper proposes a hybrid MPA-GSO-DNN model based on manufacturing service to address the personalized recommendation problem for service composition. Firstly, a hybrid multi-objective preference analysis model and glowworm swarm optimization algorithm (MPA-GSO) is proposed to generate deep learning training set by analyzing customer preference and repetitively simulating the customer s selection process. The glowworm swarm optimization (GSO) algorithm is improved with dynamic step to solve the continuous multi-objective optimization in MPA-GSO. Secondly, a deep neural network (DNN) is structured to analyze candidate services and provide personalized recommendation. Finally, a case study is presented to demonstrate the performance and practicability of the proposed approach.