Time series analysis with graph-based semi-supervised learning

With the exponential growth of time-stamped data from social media, e-commerce and sensor systems, time series data analysis is of growing interests for extracting useful insights. In many real-world applications, there is usually a large amount of unlabeled data but limited labeled data, which can be difficult to obtain. In this paper, we present a graph-based semi-supervised learning framework which leverages the unlabeled data to improve the performance of time series classification. To effectively capture the underlying structure of time series data with graphs, we explore different time series modeling techniques, and develop a probabilistic method for learning optimal graph combination. Experimental results on real-world data show the superiority of our approach over existing methods.