Bachelor´s Degree Project Presentation: "Modeling Apartment Sale Prices with Multiple Linear Regression"

Abstract:
This thesis explores multiple linear regression models for predicting apartment sale prices. The data is collected from a relatively new building to minimize factors that might be hard to track, such as the condition of the apartment. The variables in the data set are final price, asking price, monthly fee, other app, rate month, season, floor, balcony, brokerage and gender.

Four models are then identified, analyzed, and used for prediction. The first model uses all variables, the second is found by backwards elimination using the first as its starting point. The third model is also found by backwards elimination, but its starting point is a model with all variables except asking price. Lastly, the fourth model is based on what variables I believe affects the final price. Afterwards the best model is re-evaluated on the whole data set, based on number of variables,
R2, AIC score, and MSPE.