PREDICTION IN EVOLVING DATA STREAMS USING AN ADAPTIVE SYSTEM
Abstract
Mining of data streams is flourishing with many applications operating on streaming datafrom various sources, such as meteorological data and surveillance data. Data in data streamsarrives in near real time and patterns in the stream can change or evolve over time. For datastreams, current prediction techniques are unsuitable as they use a forget and rebuildmechanism where prediction models are rebuilt when changes occur, needlessly rejectingvaluable prior information. We aim to develop an adaptive predictive system, which can beapplied across various sectors including sensor networks, and climate monitoring. We will beable to continuously capture and model knowledge within a data stream using a probabilisticnetwork to allow proactive detection of changes within the stream.The relevance of this project to ONR falls under the Naval S&T Focus area of informationdominance specifically a cross between computer science and data analytics. This researchcan be applied to areas such as sensors for power control and meteorological weathermonitoring.The PIs of the project include Dr Yun Sing Koh and Prof Gillian Dobbie. The resourcesrequired include a research programmer and a PhD student. The outcomes of this particularresearch would be two-fold: (1) an open-source predictive system for evolving data streams,(2) research publications arising from this project.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jul 08, 2020
- Source ID
- N629091912034
Entities
People
- Yun Sing Koh
Organizations
- Office of Naval Research
- United States Navy
- University of Auckland