Tracking Property Ownership Variance and Forecasting Housing Price with Machine Learning and Deep Learning
Abstract
This paper proposes a research framework for smart city planning utilizing machine learning and deep learning techniques applied to multiple big data sets on real estate. Ensemble machine learning models were constructed to track property ownership variance in Austin, TX, USA, with the Random Forest model showing superior performance. Additionally, the Long Short-Term Memory (LSTM) model was employed to forecast property values in the same area. Root mean squared errors were calculated for model validation, with LSTM parameter settings adjusted to prevent underfitting or overfitting. The Random Forest model identified key factors influencing property ownership variance, including demographic, land use, and built environment factors. The LSTM model predicted a continued rise in housing prices in Austin and pinpointed areas experiencing increases or decreases in property value. These predictive models offer valuable insights for city planners to anticipate and quantify the impacts of neighborhood development.
Team
Junfeng Jiao
Seung Jun Choi
Weijia Xu
Acknowledgments
This work has been funded through The Good System project by The University of Texas at Austin.
