Diabetes Prediction and Analysis

This project is part of a machine learning course at university and involves analyzing a diabetes dataset to build a predictive model for diabetes based on various health metrics.

Data Preprocessing and Visualization

Reading Dataset and Summary Statistics

• The dataset is read using pandas, and summary statistics are printed to understand the data distribution.
• Missing values in features (except for the target variable Diabetic) are imputed with the median value of the respective feature.

Distribution of Diabetic Class

• A Kernel Density Estimate (KDE) plot is generated to visualize the distribution of the Diabetic class.

Histograms for Number of Diabetics in Each Age Group

• A histogram is plotted to show the number of diabetics in different age groups.

Density Plots for AGE and BMI

• KDE plots are generated for the AGE and BMI features to visualize their distributions.

Visualizing Correlations

• A heatmap is created to visualize the correlations between input features and the Diabetic class.

Linear Regression Models

A linear regression models are built to predict the AGE feature based on various input features.

Models Details

• Model LR1: Uses all features except AGE to predict AGE.
• Model LR2: Uses the NPG feature to predict AGE.
• Model LR3: Uses the NPG, DIA, and PGL features to predict AGE.

Evaluation

• The Mean Squared Error (MSE) and R-squared values are calculated for each model.
• Scatter plots are generated to visualize actual vs. predicted AGE values.

k-Nearest Neighbors Classification

The k-Nearest Neighbors (kNN) classifier is used to predict whether a person is diabetic.

Model Details

• Different values of k and distance metrics (euclidean, manhattan, chebyshev) are evaluated.
• Performance metrics such as precision, recall, F1 score, accuracy, ROC/AUC score, and confusion matrix are calculated and visualized.

Evaluation

• ROC curves and confusion matrices are plotted for different values of k and metrics to analyze the model's performance.