Dimension reduction model for the combination of primary data and validation data

• 1、MOFCOM,Beijing 100710

Abstract：IThe datas used in the economic forecasting problems can be divided into two types. They are primary data and validation data. For the former, the data has a variety of structures and large noise, and the number is very large. For the latter, the data has complete structure and small error, and the number is very small.If only one kind of data was used, the important information contained in the other kind of data may be neglected. The combination of the two kinds of datas is very important. However, the combination of the two data sets can lead to the dimension disaster, which makes the existing prediction models fail in the economic forecasting. Therefore, the problem of dimension reduction in the combination of two kinds of data under the background of big data needs Review of dimension reduction model for the combination of primary data and validation datato be solved urgently. Based on the above analysis, the measurement error has been introduced in the beginning. Then the existing dimension reduction models of the combination of two kinds of datas have been summarized. These models haved been divided into the parametric method, the nonparametric method and the semiparametric method. The advantages and disadvantages of the various methods have been analyzed and future research prospects were advanced.

Keywords： Dimension Reduction Economic Forecasting Primary Data Validation Data