Proceedings of the 13th EAI International Conference on Mobile Multimedia Communications, Mobimedia 2020, 27-28 August 2020, Cyberspace

Research Article

Predicting Sales of Cross-chain Discrete Manufacturing Products Based on LSTM

Download402 downloads
  • @INPROCEEDINGS{10.4108/eai.27-8-2020.2294348,
        author={Zhicheng  Wang and Ye  Tao},
        title={Predicting Sales of Cross-chain Discrete Manufacturing Products Based on LSTM},
        proceedings={Proceedings of the 13th EAI International Conference on Mobile Multimedia Communications, Mobimedia 2020, 27-28 August 2020, Cyberspace},
        publisher={EAI},
        proceedings_a={MOBIMEDIA},
        year={2020},
        month={11},
        keywords={value chain cross-chain prediction lstm},
        doi={10.4108/eai.27-8-2020.2294348}
    }
    
  • Zhicheng Wang
    Ye Tao
    Year: 2020
    Predicting Sales of Cross-chain Discrete Manufacturing Products Based on LSTM
    MOBIMEDIA
    EAI
    DOI: 10.4108/eai.27-8-2020.2294348
Zhicheng Wang1,*, Ye Tao2
  • 1: College of Information Science and Engineering, Ocean University of China, Qingdao, Shandong, China
  • 2: Qingdao University of Science and Technology
*Contact email: wangzhicheng@stu.ouc.edu.cn

Abstract

In the mass customization product value chain, the sales forecast in the marketing chain is important. Traditional forecasting methods only consider the previous sales volume, which is a single value chain factor. In fact, with the rapid development of online media, the factors affecting the sales of manufacturing products have become increasingly complex. For example, the attention and evaluation of online social media is an important influence on sales forecasts. In addition, other influencing factors such as stocks also have a certain impact on product sales. Therefore, it is possible to integrate cross-domain information to improve the accuracy of prediction results.In this paper, we propose an LSTM model with the portal's attention index, and involve factors from multiple domains, such as the network media, market environment and original product sales volume. Experiment results show that using the proposed cross-domain approach obtains more accurate prediction results than other mainstream models.