mobile data stream mining (foundations)
TRANSCRIPT
1. Background Resource-awareness Context-awareness Screen Clutter-awareness Mobile Data Stream Mining (Foundations) Dr Mohamed Medhat Gaber Reader, School of Computing Science and Digital Media Robert Gordon University 21 May 2015 Dr Mohamed Medhat Gaber Reader, School of Computing Science and Digital Media Robert Gordon University Mobile Data Stream Mining (Foundations) 2. Background Resource-awareness Context-awareness Screen Clutter-awareness 1 Background 2 Resource-awareness 3 Context-awareness 4 Screen Clutter-awareness Dr Mohamed Medhat Gaber Reader, School of Computing Science and Digital Media Robert Gordon University Mobile Data Stream Mining (Foundations) 3. Background Resource-awareness Context-awareness Screen Clutter-awareness 1 Background 2 Resource-awareness 3 Context-awareness 4 Screen Clutter-awareness Dr Mohamed Medhat Gaber Reader, School of Computing Science and Digital Media Robert Gordon University Mobile Data Stream Mining (Foundations) 4. Background Resource-awareness Context-awareness Screen Clutter-awareness Introduction to Data Streams - Big Data The advances in data acquisition hardware, and the emergence of applications that process continuous ow of high velocity data records have led to the data stream phenomenon. A data stream is a continuous, rapid ow of data that challenge our state-of-the-art processing and communication infrastructure. Data streams have led to the emergence of Big Data The general features of data streams are: Very high rate input data Read only once by an algorithm Real time processing demand Unbounded Time varying Dr Mohamed Medhat Gaber Reader, School of Computing Science and Digital Media Robert Gordon University Mobile Data Stream Mining (Foundations) 5. Background Resource-awareness Context-awareness Screen Clutter-awareness Smartphones are really smart! With continuous advances in computational power and communication abilities for smartphones and tablet computers; and The sheer amounts of data streams that we subscribe to or acquire using the onboard sensing capabilities There is an unprecedented opportunity to perform complex data analysis tasks that can benet mobile users Dr Mohamed Medhat Gaber Reader, School of Computing Science and Digital Media Robert Gordon University Mobile Data Stream Mining (Foundations) 6. Background Resource-awareness Context-awareness Screen Clutter-awareness Internet of Things Sensors Are Everywhere Dr Mohamed Medhat Gaber Reader, School of Computing Science and Digital Media Robert Gordon University Mobile Data Stream Mining (Foundations) 7. Background Resource-awareness Context-awareness Screen Clutter-awareness Mobile Data Mining: A Denition Running data mining techniques onboard smartphones utilising the continuous increase of the processing power of these devices. Two generations for Mobile Data Mining: In a client server architecture, the mobile phone is used as a knowledge presenter (MobiMine 2002) Data mining is performed onboard the mobile phone (VEDAS 2004, OMM 2009, PDM 2010, MARS 2012, and STAR 2015) Two computational modes according to the application: Single-node (centralised) Multiple-node (distributed) Dr Mohamed Medhat Gaber Reader, School of Computing Science and Digital Media Robert Gordon University Mobile Data Stream Mining (Foundations) 8. Background Resource-awareness Context-awareness Screen Clutter-awareness Compelling Applications Environment Monitoring and Emergency/Disaster Management Enabling real-time decision making Healthcare Patient monitoring Emergency/Triage management Dr Mohamed Medhat Gaber Reader, School of Computing Science and Digital Media Robert Gordon University Mobile Data Stream Mining (Foundations) 9. Background Resource-awareness Context-awareness Screen Clutter-awareness Compelling Applications Intelligent Transportation Systems Intersection safety Safety at curves Driver vigilence decline detection Real-Time Business Intelligence Dynamic management of courier pick up/drop o Mobile policing Task allocation in taxis Dr Mohamed Medhat Gaber Reader, School of Computing Science and Digital Media Robert Gordon University Mobile Data Stream Mining (Foundations) 10. Background Resource-awareness Context-awareness Screen Clutter-awareness Challenges for Data Mining in Mobile/Embedded Environments Data as a continuous stream Resource constraints Iterative nature of learning algorithms Application constraints Real-time decision making needs Intermittent connectivity Dierent applications, dierent types of analysis Dr Mohamed Medhat Gaber Reader, School of Computing Science and Digital Media Robert Gordon University Mobile Data Stream Mining (Foundations) 11. Background Resource-awareness Context-awareness Screen Clutter-awareness Data Stream Mining in Mobile/Pervasive Environments Cost-ecient, Intelligent and Real-time Data Stream Mining techniques that can: adapt to the context of diverse applications cope with and leverage distributed computational platforms take into account available resources take into account presentation needs Dr Mohamed Medhat Gaber Reader, School of Computing Science and Digital Media Robert Gordon University Mobile Data Stream Mining (Foundations) 12. Background Resource-awareness Context-awareness Screen Clutter-awareness Systems and Architectures - for Mobile/Embedded Data Stream Mining MobiMine @ UMBC VEDAS Vehicle Data Stream Mining @ UMBC / Agnik Situation-Aware Adaptive Data Stream Processing @ Monash University + Collaborators Resource-awareness Context-awareness Screen clutter-awareness Pocket Data Mining @ University of Portsmouth Distributed Mobile agent-based Dr Mohamed Medhat Gaber Reader, School of Computing Science and Digital Media Robert Gordon University Mobile Data Stream Mining (Foundations) 13. Background Resource-awareness Context-awareness Screen Clutter-awareness Situation-awareness The Big Picture Dr Mohamed Medhat Gaber Reader, School of Computing Science and Digital Media Robert Gordon University Mobile Data Stream Mining (Foundations) 14. Background Resource-awareness Context-awareness Screen Clutter-awareness 1 Background 2 Resource-awareness 3 Context-awareness 4 Screen Clutter-awareness Dr Mohamed Medhat Gaber Reader, School of Computing Science and Digital Media Robert Gordon University Mobile Data Stream Mining (Foundations) 15. Background Resource-awareness Context-awareness Screen Clutter-awareness Algorithm Granularity / Resource-awareness Dr Mohamed Medhat Gaber Reader, School of Computing Science and Digital Media Robert Gordon University Mobile Data Stream Mining (Foundations) 16. Background Resource-awareness Context-awareness Screen Clutter-awareness Algorithm Granularity Notation Dr Mohamed Medhat Gaber Reader, School of Computing Science and Digital Media Robert Gordon University Mobile Data Stream Mining (Foundations) 17. Background Resource-awareness Context-awareness Screen Clutter-awareness Algorithm Granularity Rule The Master Rule for Algorithm Granularity Dr Mohamed Medhat Gaber Reader, School of Computing Science and Digital Media Robert Gordon University Mobile Data Stream Mining (Foundations) 18. Background Resource-awareness Context-awareness Screen Clutter-awareness Algorithm Granularity Procedure Dr Mohamed Medhat Gaber Reader, School of Computing Science and Digital Media Robert Gordon University Mobile Data Stream Mining (Foundations) 19. Background Resource-awareness Context-awareness Screen Clutter-awareness AIG and AOG Demonstrated Dr Mohamed Medhat Gaber Reader, School of Computing Science and Digital Media Robert Gordon University Mobile Data Stream Mining (Foundations) 20. Background Resource-awareness Context-awareness Screen Clutter-awareness Algorithm Granularity in Action (r = memory) Dr Mohamed Medhat Gaber Reader, School of Computing Science and Digital Media Robert Gordon University Mobile Data Stream Mining (Foundations) 21. Background Resource-awareness Context-awareness Screen Clutter-awareness Algorithm Granularity in Action (r = battery) Dr Mohamed Medhat Gaber Reader, School of Computing Science and Digital Media Robert Gordon University Mobile Data Stream Mining (Foundations) 22. Background Resource-awareness Context-awareness Screen Clutter-awareness Algorithm Granularity-based Techniques Clusterers Light-Weight Clustering RA-Cluster and DRA-Cluster RA-VFKM Change Detection CHANGE-DETECT Classiers Light-Weight Class (LWClass) RA-Class and DRA-Class Frequent Items and Associations LWF (Light-Weight Frequent Items) HiCoRE (Highly Correlated Energy-Ecient Rules) Time-Series Analysis RA-SAX RA-HOT SAX Dr Mohamed Medhat Gaber Reader, School of Computing Science and Digital Media Robert Gordon University Mobile Data Stream Mining (Foundations) 23. Background Resource-awareness Context-awareness Screen Clutter-awareness RA-Cluster Example of an Algorithm Granularity-based Technique RA-Cluster is an incremental online clustering algorithm that has all the required parameters to enable resource-awareness. Memory adaptation is done through threshold adaptation and outlier and inactive cluster elimination. CPU adaptation is done through randomised assignment. Battery adaptation is done through the change in sampling rate. Dr Mohamed Medhat Gaber Reader, School of Computing Science and Digital Media Robert Gordon University Mobile Data Stream Mining (Foundations) 24. Background Resource-awareness Context-awareness Screen Clutter-awareness RA-Cluster Algorithm Dr Mohamed Medhat Gaber Reader, School of Computing Science and Digital Media Robert Gordon University Mobile Data Stream Mining (Foundations) 25. Background Resource-awareness Context-awareness Screen Clutter-awareness Memory Adaptation in RA-Cluster Dr Mohamed Medhat Gaber Reader, School of Computing Science and Digital Media Robert Gordon University Mobile Data Stream Mining (Foundations) 26. Background Resource-awareness Context-awareness Screen Clutter-awareness Battery Adaptation in RA-Cluster Dr Mohamed Medhat Gaber Reader, School of Computing Science and Digital Media Robert Gordon University Mobile Data Stream Mining (Foundations) 27. Background Resource-awareness Context-awareness Screen Clutter-awareness Processor Adaptation in RA-Cluster Dr Mohamed Medhat Gaber Reader, School of Computing Science and Digital Media Robert Gordon University Mobile Data Stream Mining (Foundations) 28. Background Resource-awareness Context-awareness Screen Clutter-awareness 1 Background 2 Resource-awareness 3 Context-awareness 4 Screen Clutter-awareness Dr Mohamed Medhat Gaber Reader, School of Computing Science and Digital Media Robert Gordon University Mobile Data Stream Mining (Foundations) 29. Background Resource-awareness Context-awareness Screen Clutter-awareness What is Context? The interrelated conditions in which something exists or occurs (Merriam Webster) The situation within which something exists or happens, and that can help explain it (Cambridge Dictionary) Any information that can be used to characterise the situation of an entity (Dey, 1999) The set of environmental states and settings that either determines an applications behaviour or in which an application event occurs and is interesting to the user (Chen, Kotz, 2000) Dr Mohamed Medhat Gaber Reader, School of Computing Science and Digital Media Robert Gordon University Mobile Data Stream Mining (Foundations) 30. Background Resource-awareness Context-awareness Screen Clutter-awareness Fuzzy Situation Inference Context-awareness Capture applications Situation Fuzzy Context Spaces Enhance probabilistic situation inferencing with fuzziness Cope with changing situations Cope with unknown situations Dr Mohamed Medhat Gaber Reader, School of Computing Science and Digital Media Robert Gordon University Mobile Data Stream Mining (Foundations) 31. Background Resource-awareness Context-awareness Screen Clutter-awareness Context Spaces The CS model provides heuristics developed specically for addressing context-awareness under uncertainty Individual signicance (i.e. weight) and contribution of context attributes in the situation space Inaccuracies of sensory originated information Characteristics of context attributes and their eect on reasoning Partial and complete containment of context-attributes values in the situation space These heuristics are integrated into reasoning formulae that are utility-based data fusion algorithms and compute the condence level in the occurrence of a situation Dr Mohamed Medhat Gaber Reader, School of Computing Science and Digital Media Robert Gordon University Mobile Data Stream Mining (Foundations) 32. Background Resource-awareness Context-awareness Screen Clutter-awareness Fuzzy Situation Inference FSI model integrates fuzzy logic principles into the Context Spaces (CS) model using the benets of fuzzy logic for modeling and reasoning about vague and uncertain situations while incorporating the CS models underlying theoretical basis for supporting context-aware and pervasive computing environments Example (CS vs. FSI) CS: SBP > 85and 135, DBP > 60and 110, HR > 45and 85 FSI: If SBP is normal and DBP is normal and HR is normal then situation is healthy Dr Mohamed Medhat Gaber Reader, School of Computing Science and Digital Media Robert Gordon University Mobile Data Stream Mining (Foundations) 33. Background Resource-awareness Context-awareness Screen Clutter-awareness Situation-awareness Dr Mohamed Medhat Gaber Reader, School of Computing Science and Digital Media Robert Gordon University Mobile Data Stream Mining (Foundations) 34. Background Resource-awareness Context-awareness Screen Clutter-awareness The Concept of Criticality Dr Mohamed Medhat Gaber Reader, School of Computing Science and Digital Media Robert Gordon University Mobile Data Stream Mining (Foundations) 35. Background Resource-awareness Context-awareness Screen Clutter-awareness Adaptation Cases Dr Mohamed Medhat Gaber Reader, School of Computing Science and Digital Media Robert Gordon University Mobile Data Stream Mining (Foundations) 36. Background Resource-awareness Context-awareness Screen Clutter-awareness Situation-aware Strategy Adjusting Parameters Dr Mohamed Medhat Gaber Reader, School of Computing Science and Digital Media Robert Gordon University Mobile Data Stream Mining (Foundations) 37. Background Resource-awareness Context-awareness Screen Clutter-awareness Integrated Strategy Adjusting Parameters Dr Mohamed Medhat Gaber Reader, School of Computing Science and Digital Media Robert Gordon University Mobile Data Stream Mining (Foundations) 38. Background Resource-awareness Context-awareness Screen Clutter-awareness Open Mobile Miner (OMM) Easy deployment of mobile data mining applications on a range of mobile devices Facilitation integration of new and existing data stream mining algorithms Interface with a range of input sources for data streams Allow exible, application specic visualisations to be developed. Dr Mohamed Medhat Gaber Reader, School of Computing Science and Digital Media Robert Gordon University Mobile Data Stream Mining (Foundations) 39. Background Resource-awareness Context-awareness Screen Clutter-awareness OMM Extensible Design Dr Mohamed Medhat Gaber Reader, School of Computing Science and Digital Media Robert Gordon University Mobile Data Stream Mining (Foundations) 40. Background Resource-awareness Context-awareness Screen Clutter-awareness OMM Demo Dr Mohamed Medhat Gaber Reader, School of Computing Science and Digital Media Robert Gordon University Mobile Data Stream Mining (Foundations) 41. Background Resource-awareness Context-awareness Screen Clutter-awareness 1 Background 2 Resource-awareness 3 Context-awareness 4 Screen Clutter-awareness Dr Mohamed Medhat Gaber Reader, School of Computing Science and Digital Media Robert Gordon University Mobile Data Stream Mining (Foundations) 42. Background Resource-awareness Context-awareness Screen Clutter-awareness Adaptive Clutter Reduction (ACR) Corollary I Corollary II Dr Mohamed Medhat Gaber Reader, School of Computing Science and Digital Media Robert Gordon University Mobile Data Stream Mining (Foundations) 43. Background Resource-awareness Context-awareness Screen Clutter-awareness Intuition Behind ACR Dr Mohamed Medhat Gaber Reader, School of Computing Science and Digital Media Robert Gordon University Mobile Data Stream Mining (Foundations) 44. Background Resource-awareness Context-awareness Screen Clutter-awareness Clutter-Aware Clustering Visualiser (CACV) Notation Dr Mohamed Medhat Gaber Reader, School of Computing Science and Digital Media Robert Gordon University Mobile Data Stream Mining (Foundations) 45. Background Resource-awareness Context-awareness Screen Clutter-awareness CACV Algorithm Dr Mohamed Medhat Gaber Reader, School of Computing Science and Digital Media Robert Gordon University Mobile Data Stream Mining (Foundations) 46. Background Resource-awareness Context-awareness Screen Clutter-awareness CACV in Action Dr Mohamed Medhat Gaber Reader, School of Computing Science and Digital Media Robert Gordon University Mobile Data Stream Mining (Foundations) 47. Background Resource-awareness Context-awareness Screen Clutter-awareness CACV Demo Dr Mohamed Medhat Gaber Reader, School of Computing Science and Digital Media Robert Gordon University Mobile Data Stream Mining (Foundations) 48. Background Resource-awareness Context-awareness Screen Clutter-awareness Interactivity iCACV Dynamic setting of visualisation thresholds Selective focusing Controlling the clusters growth Audio feedback for o-screen objects Screen fencing of clusters Dr Mohamed Medhat Gaber Reader, School of Computing Science and Digital Media Robert Gordon University Mobile Data Stream Mining (Foundations) 49. Background Resource-awareness Context-awareness Screen Clutter-awareness iCACV Demo I Dr Mohamed Medhat Gaber Reader, School of Computing Science and Digital Media Robert Gordon University Mobile Data Stream Mining (Foundations) 50. Background Resource-awareness Context-awareness Screen Clutter-awareness iCACV Demo II Dr Mohamed Medhat Gaber Reader, School of Computing Science and Digital Media Robert Gordon University Mobile Data Stream Mining (Foundations) 51. Background Resource-awareness Context-awareness Screen Clutter-awareness CACV for Stock Market Visualisation Demo Dr Mohamed Medhat Gaber Reader, School of Computing Science and Digital Media Robert Gordon University Mobile Data Stream Mining (Foundations) 52. Background Resource-awareness Context-awareness Screen Clutter-awareness Summary Mobile data mining has emerged due to technological advances and application needs. Resource constraints have been a stumbling block Developments followed the following path Resource-awareness (Algorithm Granularity) Context/Situation-awareness (Fuzzy Situation Inference) Integration between RA and SA. Adaptive Clutter Reduction (ACR) Dr Mohamed Medhat Gaber Reader, School of Computing Science and Digital Media Robert Gordon University Mobile Data Stream Mining (Foundations) 53. Background Resource-awareness Context-awareness Screen Clutter-awareness Some References Gaber, M. M., Gama, J., Krishnaswamy, S., Gomes, J. B., & Stahl, F. (2014). Data stream mining in ubiquitous environments: state-of-the-art and current directions. Wiley Interdisciplinary Reviews: Data Mining and Knowledge Discovery, 4(2), 116-138. Haghighi, P. D., Krishnaswamy, S., Zaslavsky, A., Gaber, M. M., Sinha, A., & Gillick, B. (2013). Open mobile miner: a toolkit for building situation-aware data mining applications. Journal of Organizational Computing and Electronic Commerce, 23(3), 224-248. Gaber, M. M., & Philip, S. Y. (2006). A holistic approach for resource-aware adaptive data stream mining. New Generation Computing, 25(1), 95-115. Gaber, M. M., Krishnaswamy, S., Gillick, B., AlTaiar, H., Nicoloudis, N., Liono, J., & Zaslavsky, A. (2013). Interactive self-adaptive clutter-aware visualisation for mobile data mining. Journal of Computer and System Sciences, 79(3), 369-382. Dr Mohamed Medhat Gaber Reader, School of Computing Science and Digital Media Robert Gordon University Mobile Data Stream Mining (Foundations) 54. Background Resource-awareness Context-awareness Screen Clutter-awareness Acknowledgements Prof. Arkady Zaslavsky Dr. Shonali Krishnaswamy Prof. Philip S. Yu Dr Suan Khai Chong Dr Pari Delir Haghighi and many other researcher assistants Dr Mohamed Medhat Gaber Reader, School of Computing Science and Digital Media Robert Gordon University Mobile Data Stream Mining (Foundations) 55. Background Resource-awareness Context-awareness Screen Clutter-awareness Q & A Thanks for listening! Contact Details Dr Mohamed Medhat Gaber E-mail: [email protected] Webpage: http://mohamedmgaber.weebly.com/ LinkedIn: https://www.linkedin.com/prole/view?id=21808352 Twitter: https://twitter.com/mmmgaber ResearchGate: https://www.researchgate.net/prole/Mohamed Gaber16?ev=prf highl Dr Mohamed Medhat Gaber Reader, School of Computing Science and Digital Media Robert Gordon University Mobile Data Stream Mining (Foundations)