Akademik Veri Yönetim Sistemi
6th International Scientific Research Congress, Şanlıurfa, Türkiye, 1 - 03 Kasım 2019, cilt.1, sa.1, ss.107-119
The
effects of transportation systems on its vicinity has been a popular
investigation topic. The main reason behind that popularity is that the
population of cities have been increased rapidly and caused a need for either
new living spaces or new transportation systems. The analysis of the effects of
new transportation projects in the vicinity is important for guiding to the
policy makers at implementation of proper transportation systems. In this study
a model which predict the real estate prices is established by using the direct
distances of real estate properties to the closest BRT stations and the
neighborhood characteristics. The performances of the established models are
compared by using linear regression and random forest algorithm (RFA). The closer
real estate properties to the BRT stations are expected to have higher prices.
However, the correlation test performed before the analysis have shown a
directly opposite result from the expectations. This output supports the
existence of other external factor that might have an effect on the real estate
prices rather than the direct distances to the BRT stations. Due to the prices
of the real estate properties and their direct distances to the BRT stations
have a weak and nonlinear relationship, as the direct distance between the real
estate property and the closest BRT station increases the performance of the
linear regression model decreases. On the other hand, the model that is created
by RFA shows a more irregular performance. When the dataset is limited to a
real estates that are at most 2000 meters away from the closest BRT line, RFA
shows its best performance at predictions, however when the dataset is limited
to real estates that are at most 2900 meters away from the closest BRT station,
the RFA model shows its worst performance at predictions. In future studies,
different types of data mining methods (Support Vector Regression, SVR;
Gradient Boosting, GB; etc.) and more detailed datasets will be used in order
to establish the predictive models