Modeling Dynamic Substate Chains among Massive States for Prediction

Along the development of ubiquitous sensing technologies, the opportunity to have transaction time series data is increasing. We propose a novel framework named HISC (HIgh-order Substate Chain) modeling to predict the dynamic system behaviors based on the transaction time series which contains explosive states due to the combinatorics of massive sensors and their output values with noise. Its significant performance has been confirmed through the comparisons with high-order Markov chain models and the application to practical data analysis