Abstract
The Multiple Disease Prediction System (MDPS) employs machine learning techniques—specifically Support Vector Machines (SVM) and Logistic Regression—to detect various diseases within a single platform. This system is implemented using the Django web framework, offering a user-centric interface for improved accessibility. The primary goal is to support early disease detection and encourage personalized healthcare solutions. The system focuses on predicting diabetes, heart disease, and Parkinson’s disease based on health indicators such as cholesterol levels, heart rate, pulse rate, and blood pressure. Through the use of accurate and efficient ML models, this approach identifies risk factors for multiple illnesses simultaneously. While individual disease prediction models are common, this study presents an integrated solution to predict several diseases at once. The proposed system demonstrates the effectiveness of machine learning in enhancing diagnostic capabilities and promoting better health outcomes.
Introduction
In the current healthcare landscape, diagnostic machine learning models often operate in isolation. While specialized models for conditions like diabetes or cardiovascular disease exist, they are frequently fragmented, requiring clinicians and patients to navigate multiple platforms to gain a holistic view of a patient's health. This "fragmented analysis" is inefficient, particularly as chronic conditions often share overlapping risk factors or coexist in the same patient.
To address this, we propose the Multiple Disease Prediction System (MDPS). This unified web-based platform allows users to input various health metrics—including glucose levels, blood pressure, and heart rate—to receive a comprehensive risk assessment for three major chronic conditions: Diabetes, Heart Disease, and Parkinson’s disease. By centralizing these diagnostics, the system aims to reduce the financial and cognitive burden on both patients and healthcare providers.
Methodology
The MDPS architecture follows a rigorous machine learning pipeline integrated into a functional web application. The workflow is designed to ensure data integrity and model reliability from collection through to deployment.
1. Algorithm Selection
The system utilizes two primary algorithms based on the nature of the target diseases:
- Support Vector Machines (SVM): Chosen for its effectiveness in handling high-dimensional data and defining optimal decision boundaries. This was particularly crucial for the multifaceted indicators associated with Parkinson's and Diabetes.
- Logistic Regression: Employed for binary classification tasks, proving robust for the categorical nature of heart disease indicators.
2. Technical Stack
The application is built using the Django web framework to provide a responsive and user-friendly interface. The backend processing utilizes the standard Python data science stack:
- Pandas and NumPy: For dataset management and numerical operations.
- Scikit-learn: For model training, evaluation, and data splitting.
- Pickle: Used for model serialization, allowing pre-trained models to be deployed without the need for real-time retraining upon user request.
3. Workflow Stages
The development process involved five distinct stages: Data Collection (sourced primarily from Kaggle and public health records), Preprocessing (cleaning and handling missing values), Model Selection, Data Splitting (Training/Testing), and Deployment.
Experimental Results
The proposed system demonstrated significant performance gains over existing single-disease models. The integration of specific preprocessing techniques and optimized algorithm matching led to a notable increase in diagnostic accuracy.
| Disease | Algorithm | Existing Accuracy | Proposed Accuracy |
|---|---|---|---|
| Diabetes | SVM Classifier | 76% | 79% |
| Heart Disease | Logistic Regression | 80% | 85% |
| Parkinson’s | SVM Classifier | 72% | 89% |
The most significant result was the 17% improvement in Parkinson’s disease prediction. This suggests that the SVM implementation was highly effective at capturing the complex patterns within the Parkinson’s dataset compared to previous benchmarks.
Discussion and Future Work
The MDPS highlights the power of integration in medical diagnostics. Beyond accuracy, the system’s primary innovation is its accessibility. By including "acceptable input ranges" and data validation alerts within the Django interface, the system bridges the gap between complex data science and practical healthcare tools.
However, several areas for future development remain:
- Expansion of Scope: Future iterations could include a wider array of conditions, such as various cancers or dermatological diseases.
- EHR Integration: Transitioning from static datasets to real-time Electronic Health Record (EHR) integration would be necessary for clinical-grade deployment.
- Global Generalization: Further testing is required to ensure the models generalize effectively across diverse global populations and demographics.
Conclusion
The Multiple Disease Prediction System represents a step toward more holistic, accessible, and proactive healthcare. By leveraging SVM and Logistic Regression within a unified framework, the system provides a scalable model for early disease detection, potentially improving patient outcomes through timely intervention.
Summary updated