Evolutionary ensemble learning method for classification problems

Abstract

The invention discloses an evolutionary ensemble learning method for classification problems, and belongs to the field of machine learning and data mining. The method comprises the following steps: firstly, generating a group of base learner groups based on tree coding and a corresponding integrator group based on binary coding for a classification problem; then, performing multi-objective evolutionary search of a base learner group and combined optimization search of an ensemble learner group; two levels of evolutionary search processes are combined and mutually fed back in information, so that a high-quality base learner group is constructed, and a high-accuracy calculation target of integrated learner individuals is generated. According to the method, two levels of search optimization processes can be dynamically combined through an evolutionary iteration mechanism, and then collaborative optimization is realized through information exchange between the search optimization processes, so that a base learner group is promoted to evolve continuously, and balance between high quality of individuals and diversity of the group is kept; Meanwhile, global optimization of the integrated learner structure is realized.

Description

technical field [0001] The invention belongs to the field of machine learning and data mining, and specifically includes a base learner search group based on tree coding and an integrated learner search group based on binary code. The two levels of search optimization processes can be dynamically combined through the evolutionary iterative mechanism, and then the collaborative optimization can be realized through the information exchange between each other. This architecture can promote the continuous evolution of the base learner population and maintain a balance between the high quality of individuals and the diversity of the population, while achieving global optimization of the ensemble learner structure. In addition, this computing architecture is beneficial to reduce the dependence of the construction process of ensemble learners on human design decisions, and instead realize the automatic generation of ensemble learners through data-driven models. Background technique...

AI Technical Summary

Problems solved by technology

[0004] Aiming at the problems existing in the existing ensemble learning, this paper invents an automatic ensemble learning method with dual evolution architecture for classification problems, by establishing the evolutionary search of the base learner population and the ensemble learner The two optimization processes of group evolution enable the continuous interaction and collaborative construction of the integrator group and the base learner group to generate an integrated model

Images