Early Detection of Parkinson's Disease Through Motor and Non-Motor Symptoms Using Random Forest Algorithm

Parkinson’s disease (PD) is a progressive neurological disorder characterized by a range of motor and non-motor symptoms. Traditional diagnostic approaches often emphasize motor impairments such as bradykinesia, rigidity, and tremors, potentially delaying early detection and intervention. This study proposes a machine learning-based diagnostic framework that integrates both motor and non-motor symptoms to improve the accuracy and timeliness of PD diagnosis. Utilizing a customized big parkinson’s disease dataset (CBPDD) combining publicly available and clinical data sources, the system incorporates advanced preprocessing techniques to handle missing values, outliers, and high-dimensional data. A random forest (RF) classifier is employed due to its robustness and interpretability, trained on features extracted from demographic, clinical, and sensor data. The model is validated using cross-validation strategies and evaluated through accuracy, sensitivity, specificity, and ROC-AUC metrics. Results demonstrate that the proposed system can reliably distinguish between PD and non-PD cases, enhancing diagnostic precision. Furthermore, a Streamlit-based application is developed to facilitate real-time deployment in clinical environments. This integrated approach offers a scalable and effective solution for early PD detection, ultimately improving patient care and resource management.