ABSTRACT
This project demonstrates the implementation and comparison of multiple machine learning models for the classification task of predicting Parkinson's disease status using a set of features extracted from voice recordings. The dataset used contains various voice features such as jitter, fundamental frequency, and relative average perturbation (RPDE). Four different classification models are trained and evaluated: Decision Tree (DT), Random Forest (RF), XGBoost (XGB), and K-Nearest Neighbors (KNN). The workflow begins with the preprocessing and selection of relevant features and target variable. The dataset is then split into training and testing subsets to evaluate model performance. Each model is trained using the training data and tested on the test data, followed by calculating the accuracy of each model. The models are then saved for future use, and their performance is compared using a bar chart visualization.The Decision Tree model, Random Forest model, XGBoost model, and K-Nearest Neighbors model are evaluated for classification accuracy. The resulting comparison of training accuracies helps determine which model performs the best for this specific task. All models are saved using the `joblib` library for future use. This approach provides a comprehensive way of training multiple models, evaluating their performance, and visualizing the results to make an informed decision about which model to deploy. The Python libraries used include Scikit-learn for model training and evaluation, XGBoost for advanced boosting algorithms, and Matplotlib for visualizing the results.Decision Tree (DT): 93.3%, Random Forest (RF): 96.4%, XGBoost (XGB): 96.8%, K-Nearest Neighbors (KNN): 92.1%. Based on these accuracy results, the XGBoost model achieved the highest accuracy of 96.8%, followed closely by the Random Forest model at 96.4%, showing excellent performance for predicting Parkinson's disease status. The models are saved using the joblib library for future deployment. This approach provides a comprehensive way to train, evaluate, and compare models, ultimately helping in the decision-making process for deploying the most effective model in a clinical or diagnostic setting.

OBJECTIVE
1. Accurate Diagnosis of Parkinson’s Disease:
Develop machine learning models that can accurately classify individuals with Parkinson’s disease (PD) from healthy controls, leveraging data such as motor function tests, speech patterns, and sensor readings. Use clinical data (e.g., motor function tests, speech data, or imaging), sensor data (e.g., wearable sensors tracking movement), or other biomarkers to train the model.
2. Early Detection of Parkinson’s Disease:
Use machine learning techniques to detect early signs of PD before symptoms become clinically evident, helping to improve early intervention and treatment outcomes. Use subtle features from speech patterns, gait analysis, or non-motor symptoms like sleep disturbances that may signal the onset of PD.
3. Prediction of Disease Progression:
Create models that can predict the rate of disease progression in individual patients, enabling personalized treatment plans and better long-term care management. Provide insights into how fast the disease will progress in a particular patient, helping to personalize treatment plans.
4. Continuous Monitoring and Symptom Tracking:
Implement real-time monitoring using wearable devices or home-based sensors, combined with machine learning, to track motor and non-motor symptoms, providing continuous insights into the patient’s condition. Develop models to automate the evaluation of PD severity using various biomarkers (speech, gait, or motor activity).
5. Improved Disease Severity Assessment:
Automate the evaluation of Parkinson’s disease severity (e.g., UPDRS scoring) through machine learning models that analyze sensor data, speech, and motor performance, offering more consistent and frequent assessments.