Knowledge extraction method fusing genetic algorithm and decision tree algorithm

Abstract

The invention discloses a knowledge extraction method fusing a genetic algorithm and a decision tree algorithm, and relates to the fields of classification, data mining and the like based on remote sensing images. The method comprises the steps of preparing a remote sensing classification result and classification feature data; randomly extracting a training sample set in a layered manner; initializing a genetic algorithm, constructing a decision tree through randomly generated gene combinations and samples, and simulating random features and random sample features of a random forest; screening a rule chain of an interested category for any constructed decision tree, and obtaining a classification result according to a rule; comparing the screened classification result with the existing classification result, calculating an error rate as the fitness of an evaluation function, and recording a rule chain and a corresponding error rate; and iterating the genetic algorithm to a specified number of times or meeting a convergence condition, sorting a series of rules in an ascending order according to error rates, and taking the rule with the minimum error rate as explicit knowledge. According to the method, implicit knowledge can be effectively converted into understandable explicit knowledge, and certain repeatability and robustness are achieved.

Description

technical field [0001] The invention relates to a knowledge extraction method combining a genetic algorithm and a decision tree algorithm, and relates to the technical fields of remote sensing image-based classification, data mining, and the like. Background technique [0002] With the rise of cloud platforms such as Google Earth Engine, remote sensing classification results based on different data and algorithms are increasing. In the application of these remote sensing classification results, the existing methods mostly select samples from these results, and continue to generate new classification results based on algorithms with black box properties. In this process, classification-related knowledge is implicitly transferred to new classifications through samples, and people cannot acquire, understand and apply this knowledge, thus hindering the progress of cognition. [0003] Among these algorithms with black-box properties, decision tree classification is close to a si...

AI Technical Summary

Problems solved by technology

[0005] In order to solve the problem of obtaining understandable explicit knowledge from existing classification results, the present invention provides a knowledge extraction method that combines genetic algorithm and decision tree algorithm

Images