Genetic Algorithms with diversity measures for the HTA diagnostic in children
Date:
Hypertension is becoming an increasingly common concern in modern society, particularly due to its recognition as a major cardiovascular risk factor. However, this concern is not always extended to children, despite the fact that cardiovascular disease prevention should begin during childhood rather than adulthood.
Classifier combination remains an active research area in machine learning and pattern recognition. Numerous theoretical and empirical studies have demonstrated that combining classifiers often produces better results than relying on individual models alone. A key challenge in classifier combination is ensuring diversity among the classifiers, which can be evaluated using statistical diversity measures.
Genetic Algorithms play an important role as search techniques for solving complex optimization problems across different application domains. Inspired by biological evolution and natural selection, these algorithms operate on populations of candidate solutions encoded as binary chromosomes and apply three main operators: selection, crossover, and mutation.
This work presents several diversity measures and introduces a Genetic Algorithm variant designed to identify, from a large number of possible classifier combinations, the subset that simultaneously maximizes classifier diversity and multiclassifier accuracy. The proposed method is further applied to predicting the risk of hypertension in children.