A Transparent Rule-Based System For Parkinson'S Disease Management

Abstract

Modern living styles and a failure to maintain one's health can cause a person to develop the medical problem known as Parkinson's disease. A little region of the brain that controls a person's posture, movement, and emotions is affected by this disorder. Although PD typically affects the elderly, it has evolved over time into a chronic condition that can affect anyone. Numerous studies have been conducted to control the disease in light of its severity, and it would be a significant accomplishment if an expert system could pinpoint the main causes of PD. As a result, this paper suggests a system of experts called the Risk Factors Analysis for PD (RFA-PD), using the rules produced by a decision tree to find the main risk factors for PD. RFA-PD encompasses of 5 stages: Class-Balance, Generate-Rules, SelectRule, Prune-Rule and Identify-Risk Factor. The positive class to negative class ratio in the PD data may be highly skewed, necessitating the adoption of class imbalance remedies. In the Generate-Rules stage, decision rules are derived from a decision tree, and in the Select-Rule stage, transparent decision rules are selected. In the subsequent Prune-Rule stage, the unnecessary and ineffective rules from the rule set are removed and consolidated into a single rule. The primary cause(s) of PD are determined in the final phase. The testing is based on the UCI Parkinson's Disease Speech dataset. © 2022 IEEE.