Mining Developing Trends of Dynamic Spatiotemporal Data Streams

Abstract

This paper presents an efficient modeling technique for data streams in a dynamic spatiotemporal environment and its suitability for mining developing trends. The streaming data are modeled using a data structure that interleaves a semi-unsupervised clustering algorithm with a dynamic Markov chain. The granularity of the clusters is calibrated using global constraints inherent to the data streams. Novel operations are proposed for identifying developing trends. These operations include deleting obsolete events using a sliding window scheme and identifying emerging events based on a scoring scheme derived from the synopsis obtained from the modeling process. The proposed technique is incremental, scalable, adaptive, and suitable for online processing. Algorithm analysis and experiments demonstrate the efficiency and effectiveness of the proposed technique.