Housing prices using Machinelearning

Specialeforsvar: Frederik Heedahl Baadsgaard

Titel: Housing prices using Machinelearning
Modeling "ejerlejligheder" in Copenhagen with data from SKAT

Abstract: This master’s thesis is treating the subject of modeling housing prices in Copenhagen and Fredericksburg using machine learning methods. The data used in this thesis is provided by SKAT. All data spans over the zip code 1001 to 1799 and
only involves ”ejerlejligheder.” After merging and cleaning the data, four models are built to estimate the values of the ”ejerlejligheder”: a linear model, a loglinear model, a random forest model, and a Black box model. We see that with an increase
in complexity of the models, the performance also increases, with the last one performing with a Mean Absolute Percentile Error of 20.65%. We then find that if we build a model that predicts whether or not an ”ejerlejlighed” is suitable for modeling, we can increase the performance of all 4 models and lessen the gap between them.

Vejleder: Rolf Poulsen
Censor:   Kim Christensen, Aarhus Universitet