A Cluster-Based Hidden Markov Model for High-Level State Discovery from Time Series
Hao Luo, Zhenyu Wu, Xinning Zhu · 2017
In order to learn the working mode of a complex system, we need to analyze observed sequential data to find the internal states of a system. In our paper, we focus on finding some fixed patterns from the sensor data, and we also want to extract the internal states and transition of the working state from original data, which has some advantages to reveal the working dynamics of the system. Due to the existence of a large number of unlabeled data, we focus on the unsupervised cluster algorithm to convert data stream into a stream of symbols to obtain the preliminary states of the system. We also propose a kind of Cluster-based Hidden Markov Model (CHMM) to approach the system dynamics (general situations) and intend to solve the predict problem based on states transition result. Finally, we validate the feasibility and effectiveness of the proposed method using two real data sets, a smart home energy disaggregation data set and a turbofan engine degradation simulation data set respectively.