Elian Angius
Vacation Rental Valuator
Appraises properties relative to their markets — trained on-demand allowing for what-if competitive analysis.
- Simulation
- Explainability
- Ranking
Overview
Hosts, investors & platforms need to know what a short-term rental is worth — not just an absolute nightly rate, but how it ranks against comparable listings nearby. This project is an automated valuation model (AVM) that appraises a target property — using attributes like location, room counts, images, amenities & building type — & ranks its relative value within its market.
The Challenge
- High-end & low-end markets behave differently from one another
- Markets drift seasonally & move at different velocities
- A single one-model-fits-all approach was too erroneous
- Training one model per region was too constrained to scale
- Perpetually retraining every regional model was too expensive & wasteful
Approach
- Deployed as an on-demand-training, lightweight random forest model
- Users select a custom market area of their own choosing
- The model retrains on the latest available data on demand, avoiding drift from stale training sets
Results
- Used SHAP to explain the relative contribution of each feature to a property’s value — e.g., $/bedroom/day
- Enabled what-if simulations: “what if I added a hot tub?”, “repurposed a bedroom as a game room?”, or “renovated the kitchen?”
- Simulated supply/demand pressure by flooding or removing market competition