Splicing: System for the construction of classifier’s ensembles using Genetic Algorithms and diversity measures (2014)
This report examines the application of advanced classification techniques to improve decision making in complex problem domains, with a focus on ensemble that combine the outputs of multiple individual classifiers. The study draws on expert knowledge in machine learning, statistical analysis, and evolutionary computation to analyze the role of classifier diversity as a key factor in achieving performance gains. It presents and evaluates statistical diversity measures and introduces a genetic algorithm based approach to construct diverse and accurate classifier combinations, inspired by principles of natural evolution. Through experimental analysis, the work investigates the relationship between diversity and classification accuracy and demonstrates the effectiveness of the proposed method in a bioinformatics application, highlighting the integration of theoretical expertise and practical algorithmic design.
