An approach for sales forecasting

• Published in: Expert systems with applications 2022-11, Vol.207, p.118043, Article 118043
• Main Authors: Gustriansyah, Rendra, Ermatita, Ermatita, Rini, Dian Palupi
• Format: Article
• Language: English
• Subjects: Data mining; Decision-making method; k-Means; RFM; Sales forecasting method
• ISSN: 0957-4174; 1873-6793
• DOI: 10.1016/j.eswa.2022.118043

•A new sales forecasting model that integrates Best-Worst and k-Means methods.•Best-Worst Method is utilized to extract significant forecasting criteria.•K-means method is applied to improve the quality of product clustering results.•The RFM model is used to assess the product. Inaccurate sales forecasting (SF) leads to over-stock or stock-out, in which can increase inventory costs, thereby reducing profits and return on investment. Furthermore, stock-out can eliminate customer loyalty, as well as opportunities to acquire new customers, and maximize sales or competitive advantage. This study proposed a retail SF model called the SalesKBR, which integrates the decision-making (Best-Worst Method/BWM) and the data mining methods (k-Means) into the Recency-Frequency-Monetary (RFM) model. The BWM was used to extract criteria that have a significant influence on retail SF. The k-Means method and six validity indices were applied to improve the quality of the product clustering results, while the RFM model was used for assessment. The extraction results from BWM that correlate with the retail company databases showed the criteria that significantly affect SF are frequency, quantity, and monetary. The results also showed that SalesKBR is a retail SF model with a reasonable level of accuracy. Therefore, it can be utilized to forecast retail sales, and it has the potential to be an alternative solution for making scientific and valuable management decisions.