A carregar...
Publicação
Airbnb Price Prediction in Lisbon and Porto: A Machine Learning Approach with Multi-Feature Integration
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 2.16 MB | Adobe PDF |
Autores
Resumo(s)
Airbnb’s dynamic and flexible pricing grants hosts the autonomy to set their prices, leading to
price variability influenced by seasonal changes, listing attributes, and overall demand. This
variability presents challenges for hosts in setting competitive prices and for guests in
assessing the true value of listings. In order to address these challenges, a multi-feature Airbnb
price prediction model incorporating listings, reviews, geospatial data, sentiment analysis, and
topic modeling was developed and specifically trained for Lisbon and Porto. Extensive feature
engineering and selection were first performed, followed by the application of machine
learning algorithms such as Random Forest Regression, XGBoost Regression, and Neural
Networks. Hyperparameter tuning was then conducted, and model performance was
evaluated using metrics like R², Adjusted R², MAE, MSE, and RMSE. The hypertuned multifeature XGBoost model for the combined dataset of Lisbon and Porto achieved the highest R²
value of 0.775, surpassing the performance of models trained on individual city datasets,
confirming that integrating a more extensive and diverse training dataset enhances the
model’s generalizability, leading to more accurate predictions on unseen data. This thesis
further demonstrated that models incorporating multiple additional features outperformed
the baseline model, illustrating the importance of extracting sentiment and topics from
reviews instead of solely relying on Airbnb’s qualitative ratings, and integrating geospatial
features, particularly proximity to the coastline and city center, proving to be of high
importance, enhancing the performance of the Airbnb price prediction model. Basic listing
features such as accommodation capacity and room type also play a crucial role in price
determination.
Descrição
Dissertation presented as the partial requirement for obtaining a Master's degree in Data Science and Advanced Analytics, specialization in Business Analytics
Palavras-chave
Airbnb Price Prediction Ensemble Machine Learning Artificial Neural Networks Geospatial Analysis Sentiment Analysis Topic Modeling SDG 11 - Sustainable cities and communities