Estimated Delivery Model
Understand how to design an accurate and scalable food delivery time estimation system by leveraging historical delivery data and Gradient Boosted Decision Trees. Explore model training, residual error correction, and iterative improvement to optimize delivery time predictions and balance under- and over-estimation effects.
We'll cover the following...
3. Model
Feature Category | Feature engineering | Description |
Order features | Subtotal, Cuisine | Basic characteristics of the order itself |
Item features | Price, Type | Information about individual items within the order |
Order type | Group, Catering | Indicates whether the order is a group order, catering request, etc. |
Merchant details | — | Merchant-specific metadata (e.g., ratings, popularity, etc.) |
Store ID | Store Embedding | Dense representation capturing latent store characteristics |
Realtime feature | Number of orders, Number of dashers, Traffic, Travel estimates | Live system load and conditions |
Time feature | Time of day (lunch/dinner), Day of week, Weekend, Holiday | Temporal context around the delivery |
Historical aggregates | Past X weeks average delivery time for: Store / City / Market / Time of Day | Rolling historical statistics to capture service performance trends |
Similarity | Average parking times, Variance in historical times | Metrics from similar past deliveries to estimate expected behavior |
Latitude / Longitude | Measure estimated driving time between restaurant and consumer | Spatial/geographic context used to estimate delivery time |
Training data
- We can use historical deliveries for the last 6 months as training data. Historical deliveries include delivery data and actual total delivery time, store data, order data, customers data, location, and parking data.
Model
Gradient Boosted Decision Tree
- Gradient Boosted Decision Tree sample
-
How do Gradient Boosted Decision Trees work?
-
Step 1: Given historical delivery, the model first calculates the average delivery time. This value will be used as a baseline.
-
Step 2: The model measures the residual (error) between prediction and actual delivery time.
...
-