🩺 Project Overview

This project focuses on developing a diabetes prediction tool using supervised machine learning algorithms. The goal was to explore how data-driven healthcare applications can help identify individuals at risk and support preventive care initiatives.

📊 Dataset & Preprocessing

I used the publicly available Pima Indians Diabetes Database, which contains medical data for female patients aged 21 and above. The dataset includes key attributes such as glucose concentration, blood pressure, BMI, and age.
Key preprocessing steps included:

• Handling missing or zero values in critical columns

• Normalizing data using feature scaling

• Splitting data into training and testing sets for fair evaluation

🤖 Model Selection & Training

Several classification algorithms were tested to compare performance:

• Logistic Regression

• Random Forest Classifier

• Support Vector Machine (SVM)

• K-Nearest Neighbors (KNN)

After tuning hyperparameters, Random Forest and Logistic Regression models performed best in terms of accuracy and balanced recall.

📈 Evaluation & Results

To evaluate the models, I used:

• Accuracy score to measure overall correctness

• Confusion matrix for a detailed look at false positives and false negatives

• Precision & Recall to ensure the model isn't biased towards majority classes

These metrics help validate the reliability of the predictions, especially important in healthcare-related models.

⚙️ Tech Stack & Tools

• Languages: Python

• Libraries: scikit-learn, NumPy, Pandas, Matplotlib, Seaborn

• Version Control: Git & GitHub

🧰 Highlights & Learnings

• Hands-on implementation of an end-to-end ML workflow from data cleaning to deployment-ready code

• Experience tuning machine learning models and analyzing trade-offs between precision and recall

• Improved understanding of applying ML to sensitive domains like healthcare where balanced evaluation matters

🔗 Project Link

View the full project on GitHub →