Market turning points forecasting using wavelet analysis

• A model is proposed to forecast major turning points of stock markets.
• Our model employs the system adaptation framework and wavelet analysis.
• This model is successfully applied to the US, UK and China stock markets.
• This model could be extended to other economic time series or financial markets.

Based on the system adaptation framework we previously proposed, a frequency domain based model is developed in this paper to forecast the major turning points of stock markets. This system adaptation framework has its internal model and adaptive filter to capture the slow and fast dynamics of the market, respectively. The residue of the internal model is found to contain rich information about the market cycles. In order to extract and restore its informative frequency components, we use wavelet multi-resolution analysis with time-varying parameters to decompose this internal residue. An empirical index is then proposed based on the recovered signals to forecast the market turning points. This index is successfully applied to US, UK and China markets, where all major turning points are well forecasted.