MACHINE LEARNING ALGORITHMS FOR FORECASTING ASSET PRICES: AN APPLICATION TO THE HOUSING MARKET
MACHINE LEARNING ALGORITHMS FOR FORECASTING ASSET PRICES: AN APPLICATION TO THE HOUSING MARKET

Summary/Abstract: This article investigates the application of advanced machine learning algorithms for forecasting housing prices. To this end we leverage a dataset of 414 observations of housing deals in Taipei and model it with both traditional econometric and novel machine learning algorithms. An exhaustive search among 107 alternative methods is conducted and their forecast accuracy is reported in detail. Using the root mean squared error (RMSE) as a benchmark metric, we find that implementations of the random forest family have superior performance, far surpassing that of more traditional approaches such as the multiple linear regression. The results have utility for both academics and practitioners and can be easily transferred to other forecasting problems in economics and business.

• Details
• Contents