Healthcare Diabetes Risk Prediction

Project Overview

This project analyzes clinical patient data to predict diabetes risk using SQL, Python (Machine Learning), and Tableau visualization.

The goal is to simulate a real-world healthcare analytics workflow including:

• Data cleaning & quality checks (SQL)
• Risk modeling using Logistic Regression (Python)
• Model evaluation (ROC, Accuracy, Classification metrics)
• Risk segmentation (Low / Medium / High)
• Interactive healthcare dashboard (Tableau)

---

Tools & Technologies

• SQL (PostgreSQL)
• Python (Pandas, Scikit-learn, Matplotlib, Seaborn)
• Machine Learning (Logistic Regression)
• Tableau (Data Visualization)
• GitHub (Project Portfolio)

---

Dataset

Clinical dataset containing patient health indicators:

• Glucose
• BMI
• Blood Pressure
• Insulin
• Age
• Diabetes Outcome

---

Machine Learning Model

Logistic Regression was used to predict diabetes probability.

Model Performance

• Accuracy: ~0.80
• ROC AUC: ~0.85
• High Risk Precision: strong clinical signal

The model was further used to create Risk Segments:

• Low Risk
• Medium Risk
• High Risk

---

SQL Analysis

• Data Quality Check
• Diabetes prevalence
• Feature importance (avg by outcome)
• Risk segment distribution
• Probability sanity check

---

Tableau Dashboard

The dashboard visualizes:

• Diabetes prevalence
• Risk distribution
• Probability vs Outcome
• Feature comparison
• Clinical insights
• Tableau Public link: https://public.tableau.com/app/profile/nikolaos.giannoulis/viz/HealthcareDiabetesRiskPredictionMachineLearningClinicalAnalytics/DiabetesRiskPredictionClinicalAnalyticsDashboard

---

Business / Healthcare Insight

• High glucose and BMI strongly correlate with diabetes
• Risk segmentation can help early screening
• Model can support preventive healthcare analytics