🧠 Pima Diabetes Prediction

This repository demonstrates an end-to-end machine learning pipeline for binary classification using the Pima Indians Diabetes dataset. It applies both a tuned neural network and XGBoost classifier, along with data preprocessing, exploratory data analysis, outlier handling, and visualization.

---

🔍 Problem Statement

Diabetes is a chronic condition with significant public health impact. The goal is to accurately classify whether a patient has diabetes based on medical measurements.

---

📊 Dataset

• Source: UCI Machine Learning Repository
• Observations: 768 patients
• Features:
  • Pregnancies
  • Glucose
  • BloodPressure
  • SkinThickness
  • Insulin
  • BMI
  • DiabetesPedigreeFunction
  • Age
  • Outcome (target variable: 1 for diabetic, 0 for non-diabetic)

---

💡 Key Features of the Project

✔️ Exploratory data analysis with distribution plots and correlation matrix
✔️ Outlier detection using the IQR method
✔️ Missing/impossible values imputed with medians
✔️ Data standardization
✔️ Deep Learning (Neural Network with keras-tuner for optimization)
✔️ Tree-Based Model (XGBoost) for benchmarking
✔️ ROC/AUC comparisons and confusion matrix visualizations
✔️ Fully modular and reproducible codebase

---

🧪 Models Implemented

Model	Library	Notes
Neural Network	TensorFlow/Keras	Tuned using keras-tuner (RandomSearch)
XGBoost	XGBoost	Strong tree-based benchmark

---

🚀 How to Run

1. Clone the repository

git clone https://github.com/yourusername/pima-diabetes-prediction.git
cd pima-diabetes-prediction

Install dependencies

pip install -r requirements.txt

📊 Model Evaluation Results

🔗 Correlation Matrix

📈 Feature Distributions

📉 Accuracy and Loss Over Epochs

📉 ROC Curve Comparison