Machine Learning for Predicting the Prices of Dwellings in Small and Large Cities of Finland
Sjöblom, Santeri (2023)
Sjöblom, Santeri
2023
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi-fe2023041837522
https://urn.fi/URN:NBN:fi-fe2023041837522
Tiivistelmä
Dwellings are one of the most expensive purchases individuals make in their lifetime. Additionally, dwellings can be considered as investment assets. Therefore, it is important to be able to precisely appraise the value of a dwelling, a task which can be achieved through the use of machine learning techniques. The first part of the study is a literature review which covers dwelling market dynamics, pricing of dwellings and use of machine learning in the field. The second part of the thesis presents a study on the Finnish dwelling markets, with a focus on the development of machine learning models for predicting dwelling prices in both large and small Finnish cities. Cities that have over 100,000 residents are considered as large cities and less than 100,000 residents small cities. The research datasets are divided based on the size of the cities into three datasets, with one containing all observations, one containing observations only from large cities, and one containing observations from small cities. The study tests different machine learning algorithms with each dataset and compares the best performing models of each dataset. The results show that the XGBoost algorithm is the best performing algorithm for predicting dwelling prices in Finnish cities. Furthermore, the study found that the importance of residents having a master’s degree in a district decreases in small cities, while it is the most important feature in large cities.
Kokoelmat
- 512 Liiketaloustiede [449]