ML Take-Home Assignment: Food Delivery Time Prediction

You are given historical data about food deliveries. Your task is to build a model that predicts how long a delivery will take.

Data

Under dataset/, you can find:

• train.csv — training data with target column Time_taken(min)
• test.csv — test data without the target
• Sample_Submission.csv — example file showing the required submission format and column names (includes ID and Time_taken(min))

The dataset includes features such as delivery partner attributes, restaurant & destination coordinates, order date/time, weather conditions, traffic density, vehicle/order type, and similar fields.

Like most real-world datasets, expect missing values and noisy entries.

Task

1. Train one or more models using train.csv.
2. Use a validation approach of your choice (e.g., holdout split or cross-validation).
3. Generate predictions for every row in test.csv.

Evaluation (what you should report)

• Use MAE (Mean Absolute Error) on your validation set as the primary metric. You are encouraged to use additional metrics and graphs that would make communicating
• If you apply target transformations (e.g., log), additional metrics, or specific split strategies, explain why.

Deliverables

Please submit a single zip/folder containing:

1. Short report (PDF or Markdown, ~1–2 pages) including:
   • What you built and key assumptions
   • Validation setup + metrics
   • Summary of experiments (what you tried, what helped)
   • Error analysis (at least 2 slices; e.g., traffic density buckets, distance buckets, city, weather, time-of-day)
   • Limitations and what you’d do next with more time
2. Code (notebook and/or scripts) runnable end-to-end:
   • load → preprocess → train → validate → predict on test
   • include a README.md with run instructions
3. Predictions file
   • A CSV matching the exact format of Sample_Submission.csv (same column names and ordering), containing predictions for test.csv.
4. Dependencies
   • requirements.txt (or equivalent)
5. Tooling disclosure
   • If you used LLMs or external tools, briefly note what you used them for.

Time expectations

This should fit within ~4–8 hours of focused work. Please submit within 48 hours.