Research On The Methods And RBF Neural Network Model For Clothing Sale Prediction

With the rapid development of the economy, garment industry is faced with a more complex market environment. To make a correct managerial decision adapting to the verifying environment, the best approach is to accurately forecast the future sales. The history of sales forecasting being applied to the clothing field is not long, lack of experience in the understanding of the forecasting process, and the accuracy and applicability of forecast methods led to a lot of problems in the previous studies. In this dissertation, a systematic research was done on the forecasting of garment sales.First, based on the characteristics of the garment industry, some productive and environmental factors affecting clothing sales, as well as their indices were analyzed, which established a theoretical foundation for the data collection. Then, methods used in the classification of clothing and data were advanced and different forecasting methods were analyzed theoretically.Second, according to the requirements of clothing sales forecasting, a forecast program was constructed, in which the inaccuracy, delay, time variation and other unfavorable factors could be avoided through the loop structure. Then, an experiment was done to verify the established program, based on which the relative applicability and forecasting accuracy was analyzed. By comparing it to other forecasting methods, and discussing the practical applicability of each of them in different application environment, a forecast method selection table was advanced.Finally, because of some inherent drawbacks of the statistical models in forecasting non-linear relations, an RBF network was introduced here to forecast garment sales. This network was research to achieve in garment sales forecast, including modeling process and optimizing parameter. Based on sample clustering and principal component analysis, the outliers were removed and the data dimension was reduced. Through some simulation experiments, results were analyzed and an optimal solution was obtained. By comparing the current model with other statistical ones, it was proved that the former had a better performance than the latter on overcoming the imprecision, non-linear changes and other unfavorable impacts, as well as on forecasting accuracy. In the end, it came to a conclusion that forecasting clothing sales based on the RBF network is feasible.Applying the results obtained from this dissertation into practice, it can be expected to provide some theoretical basis for clothing companies to improve the efficiency and effects in sales forecasting.