Extension Cluster Forecasting Methods Applied Research In Tobacco Sales Forecast

Tobacco industry is one of the mainstay industries of china's national economy, and it is a large family to china's tax interest. So researching on technology of predicting sales of tobacco has strong real meaning. The thesis uses extension classified prediction model, which is a somewhat new prediction one, to predict and analyze the sales of tobacco. This trial will be helpful to related decision made by managers.Extension theory researches on rules and methods of solving conflicts from qualitative and quantitative aspect. Its theory support is matter-element and extension set, and its logical cell is matter-element. Extension classified prediction is an applied technology using extension method in prediction fields. Its basis are matter-element theory and extension mathematics, and its qualitative tool is matter-element transformation, and its quantitative tool is dependant of function. It adopts combined method of quality and quantity to predict .As special commodity monopolized by our country, tobacco's every tache including price making, production, material supply, disparting sale and retail etc is all strictly controlled by government. So its sale is more influenced by external economical situation. In traditional prediction method, generally the influence can not be confirmed and described exactly and quantitatively, so veracity of prediction is affected. While matter-element of extension theory may integrate multiple environment variables as one mater-element to analyze, and make it possible combining multiple environment variables to predict. If we. regard sales change of tobacco under different economical environment as one prediction result, then we may use cluster prediction method to predict.The thesis uses extension classified prediction, principal components analysis, and R type cluster analysis methods, confirms environment index which affects sales of tobacco, and makes some attempt on selecting reasonable representation format of index shape and confirming values of classical field and partial field by mathematical statistics, and completes index selection blanks of influencing economical environment. The thesis suggests its own method and tests it in practice about parameters confirmation and results test. On the basis of real data of economical development and tobacco sales of someplace of Yunnan Province from 1990 to 2003, the thesis uses large statistical software SAS,Java to programme and obtains prediction results. These results are in accord with real economical development situation and elastic demand of tobacco commodity. This means that using extension classified prediction method to predict tobacco sales is feasible. In the meantime its analysis results have some reference meanings to strategy making of tobacco sales, market partition, and target client group partition.