Real-Time Detection of Cluster-Based Events Using the Hidden Markov Model

Ibrahim Al Sayyid, Purdue University


With the unprecedented growth in computational power and storage capabilities, there is an increasing demand for data mining and knowledge discovery in databases. Clustering is a salient area of data mining. Clustering is used to identify close objects and is useful in numerous fields such as health, economics, and marketing. In this thesis, we present a novel real time system for discovering events that are based on spatiotemporal clusters. Primitive events are defined as variations over time in cluster features. Cluster features include area, size, average radius, and density. The correlation between the clusters is assumed to be unknown. The event detection system uses the Hidden Markov Model for detecting the variations in the cluster features. The detection system can detect events such as emerging and shrinking clusters. Such events can be beneficial in the areas of epidemiology and crowd management. The effectiveness of the detection system is demonstrated by running various experiments on a synthetic dataset.




Ghafoor, Purdue University.

Subject Area

Computer Engineering

Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server