Predict monthly rent — Competition

Overview

The task. 3,000 training rows describe rental listings — city tier, floor area, bedrooms, distance to the centre, building age, furnished or not — with the monthly rent. Predict the rent for 1,200 test rows.

The data is synthetic, generated from a fixed rule plus noise, and we say so up front: nobody's real listings are in it, and there is a floor below which no model can go. That floor is the point — find the rule, and you will land near it.

Scoring. Root mean squared error, lower is better. While the competition is open the leaderboard scores about 30% of the test rows (the public split). At close, your selected submissions are scored on the other 70% (the private split) and that is the final standing. Probing the public leaderboard tells you nothing about the private one.

Baseline. Predicting the training-set mean for every row. Beating it is the first goal; a medal is the second.

Data

Row id. Use it in your submission.

city_tier

1 = capital-city inner ring, 2 = major city, 3 = regional town.

sqm

Internal floor area in square metres.

bedrooms

Number of bedrooms.

distance_km

Distance to the city centre in kilometres.

age_years

Age of the building in years.

furnished

1 if let furnished, else 0.

monthly_rent

Target. Monthly rent in US dollars (train only).

Rules

One entry per person. Your handle is fixed once you join.
Up to 5 submissions a day (resets 00:00 UTC). Select up to 2 for final scoring; if you select none, your best public score is used.
Train on the training data you can download here and nothing else. No hand-labelling of test rows.
Sharing code with another entrant during the competition is fine; sharing a submission file is not, and identical files are detected.
Results are published once and are final.