ExamplePythonLinear regression

Estimating delivery times

Sample project · A small regression notebook using synthetic delivery data.

This example shows how a project and its notebook look on this site. It uses made-up data and does not represent a completed personal project.

The question

How much does delivery time change with distance? The notebook starts with a simple straight-line model, then checks its predictions on four held-out deliveries.

Approach

Twelve synthetic samples.
Eight training samples and four test samples.
Linear regression using Python’s standard library.
Mean absolute error for evaluation.

Result

The calculations and saved outputs are shown below. The small synthetic dataset makes this useful as a layout example, rather than a meaningful benchmark.

Next steps

Use real records, compare with a baseline, and add a Gradio demo once there is a model worth trying.

NOTEBOOK

Notebook

Download .ipynb ↓

Code and saved outputs, as recorded in the notebook.

Estimating delivery times

Sample project for the portfolio preview. The data below is synthetic. This is not a completed project by Oğuz.

The question is simple: can distance give us a rough estimate of delivery time? This notebook fits a small linear regression baseline using only Python’s standard library.

1. The data

Twelve made-up deliveries, with distance in kilometres and delivery time in minutes. The first eight are used for training; the last four are held out for evaluation.

Distance (km)	Time (minutes)
1	12
2	15
3	20
4	23
5	26
6	31
7	33
8	37
9	42
10	44
11	48
12	51

In [1]

from statistics import linear_regression, mean

# Synthetic data, for demonstration only.
distances = [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12]
minutes = [12, 15, 20, 23, 26, 31, 33, 37, 42, 44, 48, 51]

train_x, test_x = distances[:8], distances[8:]
train_y, test_y = minutes[:8], minutes[8:]

print(f"Training samples: {len(train_x)}")
print(f"Test samples: {len(test_x)}")

Training samples: 8
Test samples: 4

2. Fit a baseline

The model is a straight line: time = intercept + slope × distance. It is deliberately small so we can inspect each step.

In [2]

slope, intercept = linear_regression(train_x, train_y)

def predict(distance_km):
    return intercept + slope * distance_km

print(f"Intercept: {intercept:.2f} minutes")
print(f"Slope: {slope:.2f} minutes per kilometre")

Intercept: 8.50 minutes
Slope: 3.58 minutes per kilometre

3. Check the held-out deliveries

Mean absolute error (MAE) measures the average difference between predicted and recorded times, in minutes. With only four synthetic test samples, this is a demonstration of evaluation, not evidence that the model works in practice.

In [3]

predictions = [predict(distance) for distance in test_x]
mae = mean(abs(actual - predicted)
           for actual, predicted in zip(test_y, predictions))

print("Distance   Actual   Predicted")
for distance, actual, predicted in zip(test_x, test_y, predictions):
    print(f"{distance:>5} km   {actual:>4} m   {predicted:>7.1f} m")
print(f"\nMean absolute error: {mae:.2f} minutes")

Distance   Actual   Predicted
    9 km     42 m      40.8 m
   10 km     44 m      44.3 m
   11 km     48 m      47.9 m
   12 km     51 m      51.5 m

Mean absolute error: 0.54 minutes

4. Try a prediction

A delivery of 6.5 km sits within the training range. Distances far beyond that range would be less trustworthy.

In [4]

distance = 6.5
estimate = predict(distance)
print(f"Estimated time for {distance} km: {estimate:.1f} minutes")

Estimated time for 6.5 km: 31.8 minutes

What this leaves out

Distance alone ignores traffic, preparation time, weather and the delivery route. A real experiment would need actual delivery records, a larger test set, and a comparison with a simple average-time baseline.

A next step could be a Gradio form with a distance input. No live demo is connected to this sample.