Linear fitting generally adopts the following methods

The general methods of linear fitting include fitting function based on least square method, fitting function based on pyplot and fitting function based on neural network.

Linear fitting is a form of curve fitting. Let x and y be observations and y be a function of x: y = f (x; B) Curve fitting is to seek the best estimated value of parameter b and the best theoretical curve y = f (x; B). When the function y = f (x; When b) is an I-linear function about b, this curve fitting is called linear fitting.

The problem to be solved in curve fitting is to find an analytical expression suitable for the background law; Make it the best approximation or fitting in a sense, which is called fitting model; As an undetermined parameter, the model is called linear when it only appears in the middle, otherwise it is called nonlinear.

Model selection:

For a given discrete data, it is necessary to properly select the category and specific form of the function in the general model, which is the basis of the fitting effect. If the actual background law is known, that is, the dependence of dependent variables on independent variables has an empirical formula determined by expressions, the corresponding empirical formula is directly taken as the fitting model. On the contrary, the best one can be selected by choosing different basis functions in the model.

Function plays a role in testing the adaptability of the model, so it is also called test function. Another way is to include enough test functions in the model. With the help of correlation analysis and significance test in mathematical statistics, the included test functions are screened one by one or in turn to establish a more suitable model (see regression analysis).