Hybrid Ensemble optimized algorithm based on Genetic Programming for imbalanced data classification

TL;DR

This paper introduces a hybrid ensemble algorithm based on Genetic Programming to improve classification accuracy on imbalanced datasets, especially focusing on minority class prediction, outperforming existing methods.

Contribution

It proposes a novel hybrid ensemble approach that optimizes classifier selection and sampling for imbalanced data, addressing key challenges in the field.

Findings

Achieves 40-50% higher accuracy in minority class prediction.

Effectively balances training data and optimizes classifier combination.

Demonstrates superior performance on UCI datasets.

Abstract

One of the most significant current discussions in the field of data mining is classifying imbalanced data. In recent years, several ways are proposed such as algorithm level (internal) approaches, data level (external) techniques, and cost-sensitive methods. Although extensive research has been carried out on imbalanced data classification, however, several unsolved challenges remain such as no attention to the importance of samples to balance, determine the appropriate number of classifiers, and no optimization of classifiers in the combination of classifiers. The purpose of this paper is to improve the efficiency of the ensemble method in the sampling of training data sets, especially in the minority class, and to determine better basic classifiers for combining classifiers than existing methods. We proposed a hybrid ensemble algorithm based on Genetic Programming (GP) for two classes of imbalanced data classification. In this study uses historical data from UCI Machine Learning Repository to assess minority classes in imbalanced datasets. The performance of our proposed algorithm is evaluated by Rapid-miner studio v.7.5. Experimental results show the performance of the proposed method on the specified data sets in the size of the training set shows 40% and 50% better accuracy than other dimensions of the minority class prediction.