Research And Implementation Of Sales Forecasting Based On Massive Data

With the developing of the Internet,enterprises already had a huge customer information, thisinformation provide enterprises with an effective marketing approach. However, customersinformation is so large, at the initial of system construction, considering the costs and technicalproblems, device of enterprises could not have the ability to handle such a large mass of data.Because of these deficiencies of the existing database systems, resulting in enterprises have a lotof useful data, but can not the extract useful information from it. In this paper, we analyze thefeasibility of building Hadoop distributed data processing platform, and details the Hadoopplatform technologies, project structure and architecture.To meet the market demand, we must grasp the actual needs of the market, and we know theimpact of cigarette sales market factors are diverse. In this paper, we based on method of Timeseries decomposition to build the sales forecast model, and then the model was verified. thedetailed contents about research are as the following aspects:Firstly, According to the existing databases situation in Fujian Tobacco, we analyze thenecessity of constructing a database system and then illustrate overall design and systemmodules.Secondly, we analysis the situation of Fujian Tobacco marketing platform, according tothe analysis, we research the Hadoop distributed computing platform. After understanding theadvantage of the Hadoop, we analysis of the feasibility of the tobacco companies buildingHadoop platform.Finally, we process the cigarette sales data from every province to establish sales forecastingmodel. we present the model which consider the seasonal changes and long-term trend of thecigarette market, and also verified the model. The forecast model has been applied to guide theproduction and sales in the enterprise...