Research On MS-DFM Model Selection And Its Application In Industry Cycle Identificatio

The Markov-switching dynamic factor model(MS-DFM)is widely used in characterizing economic cycles and identifying turning points in the business cycle.The model consists of two layers: a dynamic factor model and a Markov-switching autoregressive model.The former is used for dimensionality reduction,while the latter is used for characterizing and identifying the cycles.Prior to applying the model,the number of factors in the dynamic factor model and the number of states in the Markov-switching autoregressive model need to be determined.However,in existing research,there has been little systematic exploration of the model selection problem for this model,and model selection is usually based on experience.Therefore,this paper proposes two model selection algorithms based on two parameter estimation methods for the MS-DFM model: the two-step method and the two-step EM method.The effectiveness of these two methods is compared and verified through numerical simulations.Finally,this paper applies the MS-DFM model to the quantitative identification and prediction of industry cycles in the Chinese stock market from January 2000 to March 2022,using daily closing price data from 3160 stocks.This method can help us discover different trends and cyclical changes within industries.By studying the various trends and cyclical changes within each industry in depth,we can more accurately predict the future development trends and changes in the industry.