Eecient Evaluation of Composite Correlations for Streaming Time Series

In applications ranging from stock trading to space mission operations, it is important to monitor the correlations among multiple streaming time series eeciently in order to make timely decisions. The challenge is that both the number of streaming time series and the number of interested correlations can be large. The straightforward way of performing the evaluation by computing the correlation value for each relevant stream pair at each time position is not eecient enough in many situations. In this paper, we introduce an eecient method for the case where we need to monitor composite correlations, i.e., conjunctions of high correlations among multiple pairs of streaming time series. We use a simple mechanism to predict the correlation values of relevant stream pairs at the next time position and rank the stream pairs carefully so that the pairs that are likely to have low correlation values are evaluated rst. We show, through experiments, that the method signiicantly reduces the total number of pairs for which we need to compute the correlation values due to the conjunctive nature of the composites.