Title: Forecasting financial cycle: Machine learning approach Authors:  Stefan Lyocsa - Slovak Academy of Sciences (Slovakia) [presenting]
Abstract: Financial cycles are assumed to reflect the dynamics and interconnectedness between the credit, housing and stock markets, which are all important components of the overall financial stability. Estimates and accurate financial cycle forecasts could be useful for sound macro-prudential policy making and investment planning. We estimate the financial cycle for Slovakia and use machine learning techniques to predict 3- and 6-month-ahead levels of the financial cycle. The prediction accuracy is compared across multiple models and driven by a set of 170 potential predictors, including indicators related to banks, financial market, monetary policy, labor market, economic activity, business and consumer confidence and calendar effects.