2017 IEEE International Conference on Data Mining Workshops (ICDMW) (2017)
New Orleans, Louisiana, USA
Nov. 18, 2017 to Nov. 21, 2017
Predictive Complex Event Processing (CEP) constitutes the next phase of CEP evolution and provides future predictive states of the partially matched complex sequences. In this paper, we demonstrate our novel predictive CEP system and show that this problem can be solved while leveraging existing data modelling, query execution and optimisation frameworks. We model the predictive detection of events over an N-dimensional historical matched sequence space. Hence, a predictive set of events can be determined by answering the range queries over the historical sequence space. In order to take advantage of range search over 1-dimensional data structures, we transform the N-dimensional space into 1-dimension using space filling z-order curve. We propose a compressed index structure to store 1- dimensional data and execute customised range query techniques. Furthermore, we propose an approximate summarisation technique, over the historical space of top-k most infrequent range queries, to cater catastrophic forgetting of older matches. Two real-world datasets are used to demonstrate the feasibility of our proposed techniques. We demonstrate that our system can efficiently predict complex events and it equips a user-friendly interface to fulfil the requirements of user-computer interaction in a real-time.
data models, data structures, human computer interaction, human factors, query processing
S. Gillani et al., "Pi-CEP: Predictive Complex Event Processing Using Range Queries over Historical Pattern Space," 2017 IEEE International Conference on Data Mining Workshops (ICDMW), New Orleans, Louisiana, USA, 2018, pp. 1166-1171.