Prediction Method Based On Bp Network Cigarette Sales Violations

With the rapid development of the computer and information technology, we can obtain and store data more quickly and economically .The quantity of data is increasing rapidly and the knowledge hidden in it is directive to our work. Data Mining can help us to discover the knowledge and direct our work efficiently.This thesis applies Data Mining to the tobacco monopoly. Through the analysis of the large amount of historical sales recorders and all kinds of data relevant with it, a model is developed based on the BP(Back-Propagation) algorithm to predict the illegal sales of tobacco. My research can offer important and scientific basis to the supervision of the tobacco monopoly.The first part is the introduction of the conception and process of Data Mining, relevant knowledge of classification and the analysis of algorithms commonly used in classification; In the second part, the thesis gives a deep analysis on the BP algorithm used in the model. The next part is about the collection and the pre-handle of data and the goal of the model is also analyzed according to the characteristics of tobacco sales data. In the last part, the model based on Artificial Neural Network is developed and better relative parameters are determined through many experiments and also the results of the experiments are analyzed.