A comparative online sales forecasting analysis: Data mining techniques. (February 2023)

• Record Type: Journal Article
• Authors: Zhang, Bo; Tseng, Ming-Lang; Qi, Lili; Guo, Yuehong; Wang, Ching-Hsin
• Is Part Of: Computers & industrial engineering. Volume 176(2023)
• Publication Date: 2023-02
• Subjects: Data mining -- Online clothing sales -- Sales forecasting -- Optimization algorithm -- Extreme learning machine; Engineering -- Data processing -- Periodicals; Industrial engineering -- Periodicals; 620.00285
• DOI: 10.1016/j.cie.2022.108935
• Languages: English
• ISSNs: 0360-8352

Abstract

Highlights: This study contributes to present the online platform model for better decisions. The gray relational model employs to mine the features and impacts on the sales. A SFO algorithm based on random disturbance strategy (SFOR) was proposed. SFOR-ELM-based online sales prediction model is for multiple scenarios. The results prove the MAPE values controlled below 5.1% and RMSE below 16.2%. Abstract: This study aims to improve the management efficiency of e-commerce platform and assists merchants on the e-commerce platforms in formulating a suitable sales plan urgently. Online sales forecasting analysis needs to be studied and shows that the management efficiency and operating income on an e-commerce platform is improved through accurate commodity sales forecasting. A novel online clothing sales forecasting model is proposed based on data mining technique. This study contributes to presenting the model references for e-commerce platform to make decisions on future sales and directions. (1) The gray correlation model was employed to mine the correlation degree between each feature and the clothing sales to select the features that have a great impact on clothing sales. (2) A sailfish optimization algorithm (SFO) algorithm with random disturbance strategy (SFOR) was proposed based on the SFO to improve the prediction effect of clothing sales. The benchmark function test results showed that the SFOR algorithm effectively avoided local extreme points. (3) The SFOR algorithm was used to solve the extreme learning machine (ELM) random parameter problem, and the SFOR-ELM-based online sales prediction model of clothing products suitable for multiple scenarios was constructed. In addition, three cases are applied to verify the SFOR-ELM-based online clothing sales forecast model. The verification results proved that SFOR-ELM achieved satisfactory prediction results, with its mean absolute percentage error values controlled below 5.1% and root mean square error values controlled below 16.2%.