Comparación de Técnicas de Aprendizaje Automático en la Identificación de la Enfermedad de Parkinson en Personas Mayores de 60 Años

Abstract

Parkinson’s disease (PD) is a neurodegenerative disorder that primarily affects individuals over the age of 60. Early detection of this disease allows for the provision of more effective treatments and optimization of patients’ quality of life. In this research, the effectiveness of three machine learning techniques—Support Vector Machine (SVM), Random Forest (RF), and Naive Bayes (NB)—in identifying PD is compared. Voice recordings, gait data, and handwriting patterns are used for evaluation, aiming to determine the accuracy and consistency of these algorithms. The results show that RF and SVM are the most effective, with an average accuracy of 94.40% and 92.30%, respectively. SVM excelled in classifying gait data, while RF demonstrated great robustness across multiple metrics, including specificity and sensitivity. Although NB is useful, it showed limitations with an average accuracy of 70.3%, affected b