The Department of Statistics & OR is organising a seminar entitle 'Using hidden Markov and Hidden Semi-Markov Models for Financial Market Phase Detection'. The seminar will be held on Friday 5 April at 12:00 in Maths & Physics Building Lab Room 602. The speaker is Mr Luke Spiteri.
Hidden Markov models (HMMs) are the time series equivalent of independent mixture models (IMMs). While the former assumes observation independence, the latter incorporates serial dependence via a latent (hidden) discrete-time Markov chain (DTMC). Consequently, standard HMMs assume that the distribution which generates an observation at a particular point in time depends on the chosen state of the latent DTMC at that point in time. Parameter estimation and state inference of HMMs will be covered in this talk.
However, HMMs have their limitations, and one such important extension is the hidden semi- Markov model (HSMM). HSMMs generalize HMMs by allowing dwell-time distributions in states to be modelled explicitly instead of relying on the geometric distribution assumption imposed by the HMM setup. Thus, state changes and state persistence are now controlled by what is called a semi-Markov chain. Different inference approaches for HSMMs, including one based on an HMM approximation of this model, will be discussed.
In the application section of this talk, we shall look at the implementation of HMMs and HSMMs with normal state-dependent distributions to model daily returns. We look at the S&P 500 Index and the Bitcoin/USD exchange rate as case studies, with the primary aim testing these models’ adequacy at detecting ‘bullish’, ‘bearish’ and ‘correction’ phases. To test the efficacy of our models to detect these phases, we shall devise a number of model-based investment strategies against the buy-and-hold benchmark.
In the final part of this talk, we shall also look at an overview of other possible extensions of HMMs and their advantages over the standard HMM.