Healthcare Analytics: Diabetes Risk Prediction Model

Overview

An end-to-end machine learning project that predicts diabetes risk using patient health indicators. The model achieves 98.4% accuracy using Random Forest Classifier, potentially enabling early diagnosis and intervention.

Dataset

• Source: National Institute of Diabetes, Digestive and Kidney Diseases (via Kaggle)
• Size: 2,768 patient records
• Features: 8 health-related numerical attributes
• Target: Binary classification (0: Healthy, 1: Diabetic)

Project Structure

├── EDA
│   ├── EDA (Exploratory data analysis).ipynb
│   └── Healthcare-Diabetes.csv
└── Native Bayes and Random Forest
    ├── Healthcare-Diabetes.csv
    └── models.ipynb

Methodology

1. Data Preprocessing

   • Outlier detection and removal
   • Feature selection based on correlation analysis
   • No missing values or duplicates found

2. Exploratory Data Analysis

   • Distribution analysis of health indicators
   • Correlation analysis between variables
   • Feature importance evaluation

3. Model Development

   • Implemented Bernoulli Naive Bayes and Random Forest classifiers
   • Model comparison and evaluation using confusion matrices
   • Random Forest achieved superior performance

Technologies Used

• Python (Pandas, NumPy)
• Scikit-learn
• Seaborn/Matplotlib

Results

• Random Forest Classifier: 98.4% accuracy
• Successfully identified key health indicators for diabetes prediction
• Created visualization tools for healthcare provider decision support