Studies On Modeling Methods And Applications Of Grey Prediction Based On Data Processing Technology

This paper mainly follows the special ideas and logic of the grey theory,focusing on expanding and optimizing of modeling methods of grey prediction technologies.Based on the data preprocessing technology,that is,the function transformation technology,the buffer operator technology and the background value reconstruction technology,this paper makes full use of the methods and means of mathematical derivation,empirical analysis,comparative study and numerical simulation,and has made some progress,the main conclusions are:(?)A kind of transform function which can improve the smoothness of the sequence is developed,and its general function form is given.At the same time,it can be proved that it can improve the smoothness of the data effectively and increase the modeling precision.Meanwhile,taking full account of the constraint that the inverse reduction error can not be extended,this paper gives and proves the search path and function form of the transformation function under this constraint condition.To explore the possibility of multivariate function transformation technology,this paper try to put forward and prove a specific form of binary function transformation.(?)This paper constructs a new form of buffer operator,which is different from the traditional ones,and gives its theoretical proof,and verifies its good characters like form uniformity,high sensitivity,high precision,high applicability and so on.(?)Based on the theoretical background of quasi-grey index,the background value of the original data sequence obeying the approximate homogeneous exponential function is reconstructed,and a new grey prediction model is developed.The data simulation test shows the adaptability of the model under different development coefficients and the characteristics that the heteroscedasticity data series can be well fitted.Meanwhile,empirical research based on social fixed assets investment also confirms the theoretical value and application value of this model.