Business Context
RideFleet, a regional car rental marketplace operating across 120 cities, wants to understand which variables most influence daily rental demand for each vehicle. The pricing and fleet teams need both accurate forecasts and interpretable driver analysis to decide where to allocate inventory and when to retire older vehicles.
Dataset
The dataset contains one row per vehicle-day over 18 months.
| Feature Group | Count | Examples |
|---|
| Temporal | 8 | date, day_of_week, month, holiday_flag, weekend_flag, booking_lead_time |
| Location | 6 | city_id, airport_flag, tourism_index, population_density, avg_income |
| Vehicle | 9 | vehicle_age_years, make, model_group, fuel_type, transmission, seat_count |
| Pricing & availability | 7 | daily_price, discount_pct, days_available_last_30d, utilization_last_7d |
| External factors | 5 | temperature, rainfall, local_event_score, fuel_price_index, competitor_price_index |
- Size: 1.2M vehicle-day records, 35 features
- Target: Daily rental demand count per vehicle (0, 1, 2, ... bookings)
- Missing data: ~8% missing in weather fields, ~3% missing in competitor pricing, sparse missingness in tourism and event signals
Success Criteria
A good solution should:
- Achieve MAE < 0.42 bookings/day on the holdout test set
- Improve over a simple historical-average baseline by at least 20% in RMSE
- Provide a ranked explanation of the most important demand drivers
Constraints
- Predictions run in a nightly batch job for ~80K active vehicles
- The business needs interpretable feature importance, not a black-box-only solution
- Retraining should be feasible weekly on standard cloud compute
Deliverables
- Train a regression model to predict daily rental demand.
- Explain which variables most impact demand, including seasonality, location, and vehicle age.
- Describe preprocessing and feature engineering choices.
- Evaluate the model with appropriate regression metrics and a time-aware validation strategy.
- Recommend how the model would be deployed and monitored in production.
