基于神经网络的烟草零售户销售假烟行为的预警模型研究

对卷烟零售户销售假烟驱动因素分析的基础上，通过对安徽某市烟草专卖局的实地调研以及所获得历史数据的系统分析，构建了基于BP神经网络的零售户销售假烟行为的预警模型，该模型综合考虑了零售户的销售行为、零售户自身的特征以及时间因素等十二个预警输入指标。实证表明所构建的模型稳健性较好，且预测准确率得到明显提高。 

Prediction model of counterfeit cigarette sales based on neural network

Sales of the counterfeit tobacco is influencing the healthy development of tobacco industry. Furthermore, it is also undermining the national fiscal revenues. Based on the analysis of retailer's motivation for counterfeit selling and comprehensive study both on the field research results from a tobacco monopoly bureau in a city of Anhui Province and the historical data, this essay aims to establish a warning model founded on the BP neural network and counterfeit selling given the consideration on twelve factors such as the retailers selling practice, their individual characteristics and time factors. The model runs a simulation making use of the factual tobacco selling data and reaches a 72% accuracy on the warning result. This model proves to be stable.