A Study on the Applications of Emerging Sequential Patterns

This article presents a study on the techniques for detecting Emerging Sequential Patterns (ESPs) and the effectiveness of predictions made by ESPs in time-stamped datasets. ESPs are sequential patterns whose frequencies increase from one time-stamp dataset to another. ESPs capture emerging trends with time in sequential datasets and they are proposed for trend prediction. This work presents a study on the effectiveness of such predictions made by ESPs. Our experimental results show that, ESPs improve patterns' re-occurrence prediction than frequent patterns, but the improvements are marginal. Further more, we note that both ESPs and frequent patterns do not fare well in predicting the continuous emergence of patterns with time. Hence, we conclude with suggestions on future works that will improve current ESPs definition to enable detect non-trivial and interesting ESPs which can help increase the precision of predicting future emerging patterns with ESPs.