prediction using semi parametric regression function )Comparative Applied Study(

الملخص

Several regression models have been developed to address prediction issues, such as the least-squares method, the kernel regression, neural network models, and Gaussian process regression, the most recent semi parametric regression models, semi parametric methods combined parametric methods and nonparametric methods .It is important in most of studies which take in their nature more progress in the procedure of accurate statistical analysis which aim  at getting estimators efficient. In this study, we proposed  semi parametric new method to improve prediction by combining the parametric regression models represented by the least squares regression method with the non-parametric regression models represented by the Gaussian process regression. The great advantage of these models is that they contain all the positive features contained in the teacher and non-teacher model and the clarity of the interaction between the components of the teacher and non-teachers, which received wide acceptance in modern medical, economic, social and scientific studies, because of its solution to the problem of incomprehensible behavior of some variables included in the study on the one hand and for flexibility Highly enjoyed by these models on the other. The quality of the proposed method was verified by applying it to realistic and generated data using simulation. This method was also compared with the least squares regression method and Gaussian process regression using the prediction accuracy measures (MSE, RMSE, MAPE) in order to reach the best way to improve the accuracy of the prediction.The comparison that the proposed method gives the best predictive accuracy and better results in order to replicate the number of preference based on the smallest value of the values of the error measures used , because of the ability of the regression curve which ideals have an appropriate and  better data representation.