Demand forecasting method based on adjacent substitution rate estimation by GA

Product substitution is a phenomenon which occurs when the product is out of stock, and it distorts the true demand for the product while reducing retailers´ stock-out losses and improving service levels at the same time, that is expanding the demand for substitutable products. Therefore, how to estimate the product substitution is a key to improve the demand forecasting accuracy. Based on this, a novel method named EASRB-GA (Estimation of an Adjacent Substitution Rate Based on Genetic Algorithm) is proposed. First, the best weights of the attributes which affect the adjacent substitution rate are identified. Next, the weighted Euclidean distance matrix is calculated. Then, the substitution rate of the products is estimated. At last, Support Vector Machine (SVM) is used to forecast the demand. The method proposed in this paper improves the disadvantage of estimating weight by experience. As compared to the other demand forecasting models, the predict precision is improved and the objectivity and robustness are both well.