Forecast System For Cigarette Sales Base On Hadoop

The Third Plenary Session of the 18 th cpc Central Committee pointed out the strategy of “Deepening the Reform” which laid emphasis on the decisive role of market in the allocation of resources. Under this background, in 2014 March, State Tobacco Monopoly Administration presented the market-oriented reform policy. 2015, the reform started as tests in some provinces. By the tests’ experience in Beijing, Tianjin, Hebei, looking for the way of predicting cigarette demand would be the first step of cigarette sales market-oriented reform testing organization in 2015.1. To determine the predict cycle. According to the requirement of industry macro regulation and control together with industrial and commercial cigarette trading, the predict cycle is suggested to be every year and every semi year.2. To improve the predicting model. District and city lever companies would be whom the demand from. predict model should be based on local condition of economic, population and social facts, considering the factors like cigarette’s sales data, stock and price changing etc.3. To make use of demand prediction. Prediction should be based on the real demand of comsuming, reflating the comprehensive of business entities.This paper concentrates on the design and realize of Forecast Information System for cigarette sales based on Hadoop. On the aspect of annual sales prediction: improvement had been made that the linear regression algorithm was raised to multiple linear regression, which considering economy, society, population factors, and five indicators such as GDP. On the other side, monthly forecast was improved from the original time series model to the prediction model based on gray model and time series model of mixed, and realized hybrid prediction model with the Java language. Prediction is notablly accurately than single gray model or mix time series model.Finally, last work of in this paper is to integrate Forecast System for cigarette sales based on Hadoop with the existing running information system. The forecast system is very friendly to existing system without interrupting its business flow.