COMPUTATIONAL INTELLIGENCE TECHNIQUES FOR DATA ANALYSIS AND PROCESSING: APPLICATION OF MACHINE LEARNING MODELS IN THE DIAGNOSIS OF GERIATRIC DISEASES

Abstract

Stroke is one of the leading causes of death and disability worldwide, requiring rapid diagnosis to minimize sequelae and improve patient prognosis. Therefore, some Machine Learning techniques have shown promising as tools to support pre-hospital triage. This work presents the implementation and evaluation of four classification models: K-Nearest Neighbors, Random Forest, eXtreme Gradient Boosting and Support Vector Machine. These models were applied to the Kaggle "Stroke Prediction" dataset, which was subjected to preprocessing, class balancing, and hyperparameter. Furthermore, the SelectKBest technique was applied to identify the most relevant variables, targeting future applications in embedded systems. The results indicated good performance across all models, with Random Forest standing out, achieving 98.9% accuracy with 12 variables and maintaining 96.9% accuracy when reduced to just four key variables (age, hypertension, heart disease, and average blood glucose). The experiments demonstrate that models can effectively support early stroke detection, enabling their integration into mobile applications or low-cost devices for rapid and reliable screening.