A Research On The Monthly Sales Trend & Forecasting Of The General Shopping Mall Industry

With the application of the time series analysis and forecasting method, the future trend of the monthly sales figures of the general shopping mall is analyzed. In accordance with the general facts of the domestic market, the three-level dummy variables are used to identify quantitatively the dependent variables so as to develop a fitted trend and residual plots to determine directly and precisely the stationary of the adjusted time series data. By means of the Box-Jenkins forecasting method and the Minitab 13.0 and the E-VIEWS 4.0 statistical-specific software, the use value of the research results is dramatically promoted.According to the complications of the data construction, the data processing at the regular intervals are adapted with the original monthly sales divided by the days of the month, which represents the time series of the monthly sales. The effects of both the soaring and the plunging monthly sales in holidays, celebrations, sales promotions and the casual incidents - the SARS1 are intercepted by creating dummy variables which make the data aptly represent the long-range trend of the monthly sales.for the averaged daily sales in holidays, celebrations & sales promotions for the averaged daily sales in the normal coursefor the averaged daily sales in the SARS period & the subsequent period for the averaged daily sales in the non - SARS periodThe values 0 and 1 are used to digitize the qualitative variables which are constructed for a fitted averaged-daily-sales regression function of the formfor averaged daily sales at normal coursefor averaged daily sales in holidays, celebrations and sales promotionsfor averaged daily sales in SARS incidentfor averaged daily sales both in SARS incident and holidays, celebrations and sales promotionsThe dummy variables are the qualitative independent variables in the regression function of the form with a time trend isRun the Regression dialog box in the Minitab 13.0, the results of the 42 sample data are =86.4% which means that the regression model is fit to the 42 observations, with the t values of the ,and 7.64 (=230640/30205) and -6.24 (25263538769),and the p-value of 0.000. The Ftest (80.50) indicates the regression is significant. Finally, the calculation of the least squares estimates is sensitive to the averaged daily sales. The regression is highly significant and generates the residuals, few of which are large, and a string of positive residuals is followed by a string of negative residuals, followed by a string of positive residuals. And through the analysis of autocorrelation, there exists twoorder autocorrelation in this time series. Consequently the BoxJenkins methodology of forecasting is used with nine iterative approach of identifying an AR(1) and AR(2) autoregressive model. The White heteroscedasticity test identify the assumed autoregressive model is linear. The Durbin-Watson statistic of 2.05, the coefficient of determination of 0.93, but the pvalue of AR(1) model >0.05, rejecting the null hypothesis, proves that the AR(1) is not an available predictor and eliminated. The ten iterative approach of identifying an AR(2) autoregressive model which produces the DW=1.75. the R =0.93, and a better estimate of t value 0.0003 than the previous 0.0033 is proved the most appropriate for the time series business forecasting and the prediction intervals at the request.The Box-Jenkins methodology is applied to analyze both the historical and the present sates data, and forecasting the serials of the Parkson Shopping Center of Hohhot. The new sale strategy based on the forecasting results is generated to reverse the adverse situations.