Machine learning · Team project

Berlin Airbnb price prediction

How accurately can listing information predict nightly prices in Berlin, and how well does the model generalize to hosts it has never seen?

View GitHub repository
Bar chart of Airbnb price predictions within 25 and 50 euros
Selected test-set performance, shown alongside the unseen-host stress test.

Within the team project, I was responsible for modeling, evaluation, and the final analytical summary. I compared multiple model families, selected metrics that remained interpretable in euros, and separated ordinary holdout performance from the harder unseen-host test.

Team project. Yazan Aljerro’s credited scope: modeling, evaluation, and final summary.

  1. 01Leakage-aware preprocessing and train/test separation
  2. 02Baseline, linear, and ridge models
  3. 03Random forest and gradient boosting
  4. 04PyTorch multilayer perceptron
  5. 05Additional split by previously unseen hosts
  • Selected log-target random forest: MAE €29.59, RMSE €49.24, and R² 0.645.
  • 63.2% of test predictions fell within €25; 84.4% fell within €50.
  • Performance weakened on unseen hosts, exposing an important generalization limit.
  • The model estimates association, not causal pricing effects.
  • Listing data can encode neighborhood, host, and platform biases.
  • Unseen-host degradation means headline holdout metrics should not be generalized uncritically.
Pythonpandasscikit-learnPyTorchRandom forestGradient boosting
Next case studyLinear models for biological data