Fuzzy Hidden Markov Chain Based Models for Time-Series Data

The hidden Markov model (HMM) has shown a remarkable capability when dealing with time series data. However, when extended to multiple sequence datasets, the alignment is of central importance. This paper presents a new modelling design framework for pairing HMM. In this framework, an interval type-II fuzzy logic (IT2FL) model is applied for multiple sequence alignment. The inference model can determine which data sequences mostly reflect the current state of the system. Unlike the mathematical computing methods, the model requires less amount of data and leads to a low computation complexity. Results show that the accuracy of the IT2FL model is better than that of an Artificial Neural Network (ANN), especially when the datasets are not large. This model can also be used for real-time multiple time series prediction and monitoring.