Research In Genetic Programming And Its Application In Symbolic Regression

Regression analysis is one of the most important tasks in measurement data processing. The meaning of symbolic regression is that we are searching for a function that closely matches an unknown expression based on a finite set of sample data, in order to analyze and forecast data. Commonly used method is providing empirical expression model and estimating the regression parameters. The fitting results are bad when the function models are hard to be provided.Genetic programming is a new technology for solving symbolic regression problem. It can solve complex regression problem because it can obtain the matched function expression by only given the data points and acceptable error.The main research contribution of the thesis can be summarizes as follows:(1) A novel genetic programming, named symbol genetic programming (SGP) based on a new encoding method has been proposed. This new encoding method absorbs the merits of many other linear genetic programmings. It codes with a simple, unrestrained string, based on its characters, multiple expressions could be contained in one individual without the increase of computation task. This method is proved to be effective and stable from the complexity analysis and experiment.(2) By introducing Cellular Automata, a new genetic programming method named Cellular Symbol Genetic Programming (CSGP) has been proposed. It has higher success rate and need less computation time. This method is proved to be effective and stable in symbolic regression from experiments.