Forecasting Housing Output with Artificial-Intelligence-Based Big Data Analytics : An Application of Natural Language Processing and Google Trends Analytics

Big Data and artificial intelligence are quickly changing economic forecasting practices. While the big data analytics is gradually applied in the construction industry, it has not been widely used for construction predictions. Although the economists start to use Google search data to predict house prices, they have not adopted advanced machine learning techniques such as natural language processing for predicting housing output.We first review a history of forecasting techniques used in the 20th and early 21th century. With artificial-intelligence-based big data analytics, the study then uses natural language processing to select key Google search phases, and predicts housing output with Google trends, macroeconomic and market conditions data within a multivariate time series model.We find that AI-based big data analytics can help to explain the short-term dynamics in the housing markets and provide some useful short-term forecasts of construction output. Meanwhile, using all related information including macroeconomic and market conditions will improve the robustness of the forecasts.Therefore, the study fills the following research gaps: Firstly, it is the first research to use natural language processing to select key phases for constructing big data indices. Secondly, it adopts big data such as Google search data for predicting construction output in the housing market