Curve Fitting Principle And Its Application Research

Curve Fitting, Curve Fitting) refers to choose the appropriate type of Curve to fit or social science experiment, the observation data in economic activities, and the Fitting Curve equation y = f(x, c) to reveal the inherent law of variable between x and y, as a result, the Curve Fitting method of research in the natural science, social science and economic activities, and other areas of the prediction and nonlinear correction and the unknown object modeling field has important theoretical significance and application value. This article main research content is as follows:(1) Three kinds of curve fitting polynomial model was established according to the different type of curve fitting model were analyzed. At the same time, build three kinds of basis function neural network structure. Fitting with experimental data as the neural network training samples; Fixed input layer to hidden layer has a weight constant is 1; Polynomial matrix constructed as excitation function of hidden layer neurons; With polynomial model parameters as the weights of the hidden layer of neural network training; Through neural network fitting error between the output and the fitting sample value cumulative value.(2) The research methods of curve fitting coefficient algorithm, by using different algorithms for different model theory research, the contrast that is suitable for a polynomial fitting optimization algorithm of the model. The method with polynomial fitting model to construct neural network topology structure, model parameters as the weights of neural network, the experimental data were collected as the reference curve fitting, using the steepest descent method and conjugate gradient method, the recursive least square method respectively to optimize the weights of neural network training for polynomial model optimization parameters.(3) Curve fitting of the applied research, through the simulation experiment mainly discuss its application in the compensation of nonlinear systems with unknown system prediction research and analysis of experimental results. Paper with negative temperature sensor nonlinear compensation of thermistor as the simulation results and the dam displacement prediction, the national electricity consumption forecast simulation experiment, the experimental data of prediction research based on algebraic polynomial curve fitting method is not only fitting error is small, fast convergence speed, and the prediction error is small.The simulation results show that compared with the traditional fitting model this paper puts forward that the method is not only obtained the higher accuracy of fitting result, and is especially suitable for short mid-term prediction, therefore, thesis research methods in the field of all kinds of short mid-term prediction has important application value.