HealthGuard-AI: Intelligent Patient Risk Assessment System 🏥 (Milestone 1)

Problem Understanding & Healthcare Use-Case

Diabetes is a chronic disease that affects millions of people globally and can lead to severe health complications if not properly managed. Early diagnosis and proactive intervention are critical. The objective of the Intelligent Patient Risk Assessment System is to build a machine learning model capable of predicting the onset of diabetes based on diagnostic measurements. By providing a quick, accessible risk assessment tool, this system empowers patients and enables healthcare professionals to identify high-risk individuals early.

Dataset

The dataset used is the Pima Indians Diabetes Database. It contains clinical data designed diagnostically to predict whether a patient has diabetes based on several physiological measurements.

Input-Output Specification

Model Inputs (Features)

The system accepts structured clinical diagnostic data including:

• Pregnancies: Number of times pregnant
• Glucose: Plasma glucose concentration (2 hours in an oral glucose tolerance test)
• Blood Pressure: Diastolic blood pressure (mm Hg)
• Skin Thickness: Triceps skinfold thickness (mm)
• Insulin: 2-Hour serum insulin (mu U/ml)
• BMI: Body mass index (weight in kg/(height in m)^2)
• Diabetes Pedigree Function: A synthesized metric scoring diabetes likelihood based on family history
• Age: Patient's age in years

System Outputs

• Risk Score: The predicted predictive probability (0-100%) indicating the likelihood of diabetes.
• Risk Category: Actionable health categorization into Low, Moderate, or High risk.
• Top Contributing Factors: The system natively extracts and explains the top 3 specific diagnostic features raising or reducing the patient's individual calculated risk in real time.

Model Performance Analysis

The selected ML model is Logistic Regression. It was deemed ideal due to its interpretability compared to opaque neural networks and clear decision boundaries on binary classification sets.

• Preprocessing: Incorrect or missing zero-values in biological features (e.g., Blood Pressure, Glucose) were imputed with sequence medians, substantially improving data integrity.
• Performance Evaluation: The model was tested on a 20% holdout test dataset. Accuracy and strictly evaluated ROC-AUC scores dynamically generated during the compilation phase show significant reliability in successfully classifying positive instances of diabetes risk versus negative ones.

Running the Application Locally

1. Ensure Python 3.8+ is installed locally.
2. Install the application's required libraries:

   pip install pandas numpy scikit-learn streamlit joblib

3. Execute the training script to fetch the .csv data, train the model, and create the necessary predictive binaries:

   python model_training.py

4. Run the interactive Streamlit user-interface:

   streamlit run app.py

5. Open your browser to the local URL provided by Streamlit to query health predictions.