Machine learning characteristic based symbolic regression GP (Genetic Programming) algorithm

Abstract

The invention provides a machine learning characteristic based symbolic regression GP (Genetic Programming) algorithm. By using a data storage mode of Weka, table data or database data can be easily converted into a ARFF (Attribute-related file format) data file; most of table data or database data can be exported to CSV (Comma-separated value)-format data, and the data is a series of comma-separated data item values; after the exporting is completed, the file is opened in a text editor, and attributes are added in the file; adding the name of a dataset to @relation tag, adding attribute information by using @attribute, and data information is added to @data and then is saved. The save it. The machine learning characteristic based symbolic regression GP algorithm is combined with a conversion algorithm, a set similarity algorithm, a minimum Hash and local sensitive hash algorithm of a tree set so as to find a new individual similarity determining algorithm and apply the algorithm to GP.

Description

technical field [0001] Specifically, the present invention is a symbolic regression GP algorithm based on machine learning features. Background technique [0002] The main research and solution problem of regression (Regression) process is to identify and analyze the mathematical relationship hidden in the experimental data. In summary, the work of industrial engineering and scientific research is to use these experimental data and regression methods to obtain a formula or model of an experimental process, and then apply it to practice. [0003] There are many methods and tools that can help researchers to obtain such symbolic models for reverse engineering, but in most cases, they are limited to linear systems, or nonlinear systems with only a few deterministic models, such as Parametric methods such as artificial neural networks can also model nonlinear systems without pre-defined models, but they cannot clearly show the internal structure of the target system. [0004] ...

AI Technical Summary

Problems solved by technology

[0003] There are many methods and tools that can help researchers to obtain such symbolic models for reverse engineering, but in most cases, they are limited to linear systems, or nonlinear systems with only a few deterministic models, such as Parametric methods such as artificial neural networks can also model nonlinear systems without pre-defined models, but they cannot clearly show the internal structure of the target system