Predict Diabetes with Linear Regression: A Step‑by‑Step Python Guide

1. Dataset Introduction

The diabetes dataset contains 442 records with 10 attributes: Age, Sex, Body Mass Index, Average Blood Pressure, and six disease progression measures (S1–S6). The target variable represents a quantitative disease progression measure after one year.

2. Using LinearRegression

The LinearRegression model from sklearn.linear_model is trained via fit(X, y), where X are the feature columns and y is the target label.

Typical usage in scikit‑learn:

After fitting, predict() generates predictions for new feature data, returning the estimated target values.

3. Linear Regression for Diabetes Prediction

The dataset is split into training and test sets: the first 422 rows are used for training, and the last 20 rows serve as the test set ( diabetes.data[:-20] for training, diabetes.data[-20:] for testing).

Training produces a regression model; predictions on the test set are obtained with predict(). The results show the model’s coefficients, intercept, residual sum of squares, and R² score.

4. Optimizing the Code

The optimized script adds explicit calculation of the slope and intercept, computes the distance from each point to the regression line, and saves the plot with higher resolution.

5. Results and Visualizations

Final output includes the model coefficients, residual sum of squares, and the R² score, followed by plots that compare true values (scatter points) with the predicted regression line.

To hide axis tick values in the plot, additional code (shown in the image below) can be added.