Forecasting for Sales at New Super Store Using Machine Learning by Commercial District Information

This study opening a new super store involves selecting an economically feasible location through screening candidate locations and estimating expected sales. Individuals or corprate businesses may have less access to sensitive information than large distribution companies, making it difficult to select an effective location. Accordingly, this study proposed a model that can predict sales of new super stores using easily accessible commercial district information. At this time, the model applied the broken-line regression method to dived the analysis object into two areas before/after the break-point, and separately estimated sales by fitting ridge regression to each area. Through this, it is expected that it will be possible to identify various relationships between sales and influencing factors.BR As a result of the analysis, a clear difference in relationship between the two areas appeared starting from the break-point, and the sales prediction accuracy through the regression model was also found to be at a good level.