The Research Of Data Mining Based Sales Forecast

Forecast, which is part of the "Budget Planning And Forecast" in Strategic Enterprise Management, has been given the same importance as business strategy and budget planning. An enterprise's business strategy determines its development focus in a period of future time. Based on these strategies, if can be combined with the companies market plane and products sales status, sales forecast can make a comprehensive analysis of all factors and be used for product sales potential estimation. Planning itself has the portrait of uncertainty, if planned without fully considering all kinds of social factors that may affect the product's sales, this may greatly increase the company's business operations risk. With this production sales forecast system, the uncertainty of enterprise business plan can be greatly reduce and turned into a kind of more reliable operational goal. Based technologies such as data warehouse, online analysis processing as well as data mining, this research's purpose is making some analysis on those very large amount of data from manufacturing enterprises, mainly to do some prediction job on product sales and the price of those products, finally use these prediction result to provide enterprises with some sort of planning support.There are three most often used prediction methods, qualitative analysis, historical mapping method and cause-result relation analyze. However, it is really difficult to find explicit casual relationship between elements and the qualitative analysis way is so subjective, both of them are difficult to realize programmatically, so using the historical mapping method for prediction it our best choice. This article proposed a new exponential smooth algorithm based data mining model to cope with volatilization problem in enterprise's products sales data, and then do some improvement and application optimization job on data mining algorithms imported in the forecast model. Compared with traditionally using ETL layer and data warehouse for data filtering and smoothing, research shows that this improved data mining model can make a greater improvement in forecast accuracy and model performance, this model basically reached the requirement for enterprise decision support.