Skip to main content

LUP Student Papers

LUND UNIVERSITY LIBRARIES

Modelling Airbnb Prices in the Maltese Islands

Camilleri, Gabriella LU (2023) DABN01 20231
Department of Economics
Department of Statistics
Abstract
The digital platform Airbnb has gained popularity in a number of countries particu-
larly in the Maltese islands. Striking a balance in setting a price that is competitive
and also renders a good profit can be a challenge. In this thesis a model is de-
veloped to predict the price of a listing in the Maltese islands for September 2022
through a machine learning approach whereby five types of models are considered.
K Nearest Neighbours sets a baseline, while linear regression, a random forest, gra-
dient boosted trees and neural networks are assessed in search of the model that is
most generalisable beyond training data. Findings from this research conclude that
gradient boosting specifically CatBoost model gives the best... (More)
The digital platform Airbnb has gained popularity in a number of countries particu-
larly in the Maltese islands. Striking a balance in setting a price that is competitive
and also renders a good profit can be a challenge. In this thesis a model is de-
veloped to predict the price of a listing in the Maltese islands for September 2022
through a machine learning approach whereby five types of models are considered.
K Nearest Neighbours sets a baseline, while linear regression, a random forest, gra-
dient boosted trees and neural networks are assessed in search of the model that is
most generalisable beyond training data. Findings from this research conclude that
gradient boosting specifically CatBoost model gives the best performance achieving
an R2 of 0.77.
Additionally the same models are re-fitted but incorporating additional walkable
distance features to carefully identified points of interest namely historical sites,
beaches, nightclubs, the capital city and bus stops. The results attained indicate
that none of of the walkable distance features heavily contribute to explain any
variance in the price of listings in the Maltese islands and only a slight improve-
ment in model performance in some of the models considered is reported. Further
to this, while retaining the additional distance features, training of the neural net-
work is leveraged by pre-training the model on data that corresponds to another
Mediterranean touristic island of Crete and a slight improvement is reported in
model performance over the model solely trained on data for Malta from an R2 of
0.66 to 0.67. This result opens a window for further research that seek to reap the
benefits of transfer learning. (Less)
Please use this url to cite or link to this publication:
author
Camilleri, Gabriella LU
supervisor
organization
course
DABN01 20231
year
type
H1 - Master's Degree (One Year)
subject
keywords
Airbnb, Maltese Islands, Machine Learning, Geospatial Data, Transfer
language
English
id
9128841
date added to LUP
2023-11-21 12:53:44
date last changed
2023-11-21 12:53:44
@misc{9128841,
  abstract     = {{The digital platform Airbnb has gained popularity in a number of countries particu-
larly in the Maltese islands. Striking a balance in setting a price that is competitive
and also renders a good profit can be a challenge. In this thesis a model is de-
veloped to predict the price of a listing in the Maltese islands for September 2022
through a machine learning approach whereby five types of models are considered.
K Nearest Neighbours sets a baseline, while linear regression, a random forest, gra-
dient boosted trees and neural networks are assessed in search of the model that is
most generalisable beyond training data. Findings from this research conclude that
gradient boosting specifically CatBoost model gives the best performance achieving
an R2 of 0.77.
Additionally the same models are re-fitted but incorporating additional walkable
distance features to carefully identified points of interest namely historical sites,
beaches, nightclubs, the capital city and bus stops. The results attained indicate
that none of of the walkable distance features heavily contribute to explain any
variance in the price of listings in the Maltese islands and only a slight improve-
ment in model performance in some of the models considered is reported. Further
to this, while retaining the additional distance features, training of the neural net-
work is leveraged by pre-training the model on data that corresponds to another
Mediterranean touristic island of Crete and a slight improvement is reported in
model performance over the model solely trained on data for Malta from an R2 of
0.66 to 0.67. This result opens a window for further research that seek to reap the
benefits of transfer learning.}},
  author       = {{Camilleri, Gabriella}},
  language     = {{eng}},
  note         = {{Student Paper}},
  title        = {{Modelling Airbnb Prices in the Maltese Islands}},
  year         = {{2023}},
}