Research On The Prediction Of Scattered Taxes Based On Data Mining

Recently, Chinese and international scholars have already done a lot of research in taxation and data mining, and the appliance of data mining in taxation has become more widespread. Scattered tax is a part of the normal tax, generally refers to the tax comes from the individual’s or collective’s temporary business, including personal labor service and other business services. The amount of single scattered tax is always small, this feature makes scattered tax got little attention in the past studies, and also lack of studies of different various types of data mining analysis methods’application on it. However, the number of the scattered taxpayers is huge, usually accounts for over eighty percent of revenue’s taxpayer number, and scattered taxpayer is widely distributed and normative poor, this feature requires a lot of manpower and resources to strengthen the taxation, thus it’s significant to improve the taxation efficiency by making use of the advantages of information technology and looking for knowledge and law by scientifically use of historical data.Data mining is a kind of applied technique used to looking for knowledge, it’s a process that extract unknown, including hidden but credible and effective contents and information from the imperfect, relatively obscured, interfered, massive, random raw data. Mining of the predictive knowledge refers to the process looking for data’s law and thus to predict the trend based on the data from different external or existing data sources, mainly historical data, by making use of certain data mining methods, the object is to decides data’s trend in next stage based on past data’s characteristics and trends.The purpose of this paper is to establish a good scattered tax forecasting model using data mining methods. On the basis of theoretical research on scattered tax and data mining, and the characteristics of scattered taxation’s process and management, paper analyzed the differences between the scattered taxation and general taxation, constructed a scattered tax analysis indicator system capable of supporting analysis and forecasts of scattered tax. On the basis of the index system of scattered tax, paper compares the two analysis methods to be adopted to the scattered tax forecasting modeling, then proposed general steps to create a tax forecasting model. Finally, practically apply the forecasting model based on the data and information of H revenue. The main innovations:1. Applying the data mining techniques in predictive analysis of the scattered tax, and built the scattered tax analysis and forecasting indicator system. Recently, there is no study of the appliance of data mining technology in scattered tax in our country, the paper analyzes the general tax analysis system, and then based on it, built a forecasting indicator system highly targeted on scattered tax according to the management, administration and accounting features of scattered taxation.2. Separately modeled the individual tax and natural person’s tax in scattered tax. Pointed out the difference of individual tax and natural person’s tax on taxation process and data characteristics, choose the most appropriate model according to their each characteristics, and then combined them to obtain a good prediction model of scattered taxes.