Hidden Semi Markov Model

Hidden Semi-Markov Models (HSMMs) are statistical models used to analyze sequential data where the duration of each state is not fixed, unlike standard Hidden Markov Models. Current research focuses on improving HSMM estimation accuracy, particularly addressing issues like overestimating the number of states and developing robust algorithms for unsupervised learning. Applications span diverse fields, including music analysis, transportation research (e.g., modeling driving patterns), and healthcare (e.g., identifying high-risk events in ICU patients), highlighting the model's versatility in uncovering hidden patterns within time-series data. The development of more interpretable HSMM variants, such as those incorporating tree-based structures, is also a significant area of ongoing investigation.