Deep Learning for Real Estate Trading

Artificial intelligence (AI) and Machine Learning techniques have been making impact on the real estate industry recently. Increasingly, several real estate companies have started to use a variety of AI techniques to optimize their property business. Ma-chine learning (ML) technology for providing support on real es-tate investment decisions, allows investigation of historical property sales data by computer algorithms to automatically predict house prices. Real estate professionals can leverage sophisticate ML techniques to analyse sales data as benchmarks and make appraisals for their home selling clients and customers. ML technologies not only make predictions or classifications, but also can assistant real estate professionals for investment purposes by providing a trading strategy. In this paper, we propose a novel machine learning model, based on a standard deep reinforcement learning (DRL) model, enhanced with a combination of two popular time series algorithms, the Gramian Angular Field (GAF) and long short-term memory (LSTM) algorithms for providing decision support on real estate trading strategy. Our goal is to explore if the proposed enhanced DRL model can make profitable trading strategy for a long-time investment, such as the real estate markets.