Diabetes Prediction using Logistic Regression
Shravani Kondibhau Wayal
Diabetes Prediction using Logistic Regression
A machine learning-based classification project to predict whether a person is diabetic or not, based on medical diagnostic data. This project aims to support early diagnosis using logistic regression and fundamental data analysis techniques.
Problem Statement
Diabetes is a growing health concern worldwide. Early detection is crucial to prevent complications such as heart disease, kidney failure, and vision loss. This project builds a machine learning model that can assist healthcare professionals in identifying high-risk individuals by analyzing health metrics such as glucose levels, BMI, and blood pressure.
Potential Use Case
This model could be deployed by healthcare startups in rural India as a mobile-based screening tool. Health workers could collect patient data on-site and use the model to instantly predict diabetes risk — even in low-resource settings with limited internet or medical infrastructure.
Dataset Overview
Name: Pima Indians Diabetes Database
Source: Kaggle Dataset Link
Description: This dataset includes several medical predictor variables and one target variable (
Outcome). All patients are females of at least 21 years of age of Pima Indian heritage.Records: 768
Features
Feature Description Pregnancies Number of pregnancies Glucose Plasma glucose concentration BloodPressure Diastolic blood pressure (mm Hg) SkinThickness Triceps skin fold thickness (mm) Insulin 2-Hour serum insulin (mu U/ml) BMI Body mass index DiabetesPedigree Diabetes pedigree function (family history) Age Age in years Outcome 0 = Non-diabetic, 1 = Diabetic
Exploratory Data Analysis (EDA) Highlights
-Strong correlation observed between glucose levels, BMI, and diabetic outcome.
-Several columns (like insulin, skin thickness) contained zero values which were treated as missing and imputed using median.
-Diabetic patients tend to have higher BMI, more pregnancies, and higher glucose levels.
Feature Engineering
-Handled missing values (zeroes) by replacing them with column medians.
-Scaled features using StandardScaler to improve model performance.
-No feature was dropped — logistic regression performed best with all predictors.
Model Overview
Algorithm: Logistic Regression
Toolkits: Python, Scikit-learn, Pandas, Matplotlib, Seaborn
Train/Test Split: 80/20
Model Evaluation
Metric Value Accuracy 77.9% Precision 77.4% Recall 62.2% F1 Score 69.0% ROC AUC Score 84.7%
Confusion Matrix
[[82 17] [21 34]]
Classification Report
ROC Curve and AUC Score
AUC Score: 0.8229568411386594
Model Interpretation
The model performs well overall but shows slightly lower recall for diabetic cases. In a real-world medical setting, false negatives (undetected diabetic cases) are riskier and need to be minimized. A future version could adjust the classification threshold or apply recall-optimized models.
Future Enhancements
-Test other models like Random Forest, Decision Tree, or XGBoost
-Hyperparameter tuning using GridSearchCV
-Use SHAP or LIME for model explainability
-Build a Streamlit app for interactive predictions
-Integrate real-time form inputs for live testing
Dependencies
Make sure you have the following Python libraries installed:
-pandas
-numpy
-matplotlib
-seaborn
-scikit-learn
-jupyter
Get Started
Open the notebook in Jupyter:
Notes
-Zero values in columns like BMI, BloodPressure, and Glucose were treated as missing and handled appropriately.
-Evaluation is based on one train/test split; performance may vary with cross-validation or different splits.
-The model can be extended to more complex classifiers like Random Forests or XGBoost.
Project Structure
Author
Shravani Wayal
Data Science & ML Enthusiast | Python | Analytics email: wayalshravani04@gmail.com
Like this project
Posted Aug 16, 2025
A ML project to predict whether a person is diabetic or not. This project aims early diagnosis using logistic regression & data analysis techniques.
