Research And Application Of Clothing Sales Forecasting System Based On Time Series

Garment industry has the character of fast sale in the production purchasing, sales, inventory, the key issue the garment enterprises have to face lies in fasting response to the needs of short sales cycle, and improving supply chain management efficiency. In this paper, in order to improve short-term clothing sales forecast for the target rationality, analyze the impact factors of the apparel industry, and propose clothing sales forecasts solutions based on time series. The main contents are as follows:First investigate and summary the influence factors in clothing sales through real research, get the application of enterprise environments multidimensional display application by studying and analyzing time series theory and prediction theory. Then the summary of applications in auto-regression model and moving average model have been shown. This paper hold the basic idea that the current sales are associated with past sales and white noise impact from past to current state, followed by what has been discussed, the idea of how much weight to occupy with past sales and white noise. By solving the parameters of past sales and white noises, it can raise the forecasting model through least squares method with parameter estimation, finally check the validity of the prediction error test of model. In system design, this paper starts from drawing ER diagrams of business data and data tables model, define the prediction analyze and overall framework of the system. The forecasting system consists of sales data management, trend analysis and warehouse management. Core prediction process comes down to as following:with the use of the R program for extracting the original business source data, determine the structure information table, merge table field structure, generate time series document of sales, from the database product, inventory and sales field logs, collection the use in the form of time series; after that,starting based on the rationality of prediction, define week sales of apparel product as forecasting goals, define the meaning of the data parameters for forecast modeling work, And make the visualization of historical sales, generate trends plots.Core prediction process comes down to the following:the use of the R program number on the business source data were extracted to determine the structure information table, merge table field structure, sales document generation time series, the database product, inventory and sales field logs extracted, collection for use in the form of time series; starting from the prediction of rationality, defined as sales of apparel products week forecast goals, define data parameters significance. Historical sales visualize, generate trends. Then this paper introduce the mechanism of eliminating seasonal factors in short-period applicable sales, to obtain free long-term trend of the time series. The long term non-stationary time series need to take step difference converting into smooth sequence. The processed time series then need to take the sequence auto correlation function stability test, After that, with the calculating of correlation function values and partial correlation function values within the confidence interval reducing into zero-order, combined with the minimum information criterion, it can obtain the determined order of the relevant influencing factors; with the use of minimizing second multiplication principle in forecasting square error between training set and real value, the square error then calculating partial derivative to define the weighting factor regression equations weighting factors and the moving average power, as known as the optimal weighting factors. Thus it could determine the regression equation prediction model. Finally, this paper focus on the validity of the effectiveness, to check the forecasting error between the real value of the test set and forecasting model, and the stability of this model; and the prediction error is subject to zero-mean normal distribution curve, it can predict future weekly sales.Based on the core prediction model developed sales forecasting system that provides visual trend analysis, and after the trial, a large garment enterprises, improve the marketing system in contradiction sales and replenishment, operational decisions for the apparel business provides effective support.