Article Preview
Top1. Introduction
The big-data (Villars, Olofson, & Eastwood, 2011) is famous for its three attributes namely Variety, Velocity and Volume. Recently, Veracity has also become an important attribute in big-data meaning ‘data with doubt’. Both Velocity and Veracity are key ingredients of event processing applications. However, Complex Event Processing (CEP) (Academic/Commercial) (Luckham, David & Brian, 1998) takes care of ‘Velocity’ but not ‘Veracity’.
Most real time monitoring applications use CEP. The various applications where CEP is used are fraud detection (Bass, 2006), stock trading (Adi, Botzer, Nechushtai, & Sharon, 2006), security and surveillance (Bhargavi & Vaidehi, 2013; He et al., 2006), business process management (Janiesch, Christian, Martin & Oliver, 2011) Wildlife habitat monitoring (Polastre, Szewczyk, Mainwaring, Culler & Anderson, 2004), health monitoring at hospitals (Yao, Wen, Chao-Hsien & Zang, 2011; Vaidehi, Bhargavi, Kirupa & Hemalatha, 2012) etc. CEP is used as fast decision making engine by most of the applications. However, these applications fail to consider precision in data while decision making.
The existing event processing engines such as Drools (Proctor & Mark, 2012), Esper (http://esper.codehaus.org), etc. provides framework for the development of application specific rules to detect abnormalities. Moreover, existing CEP systems considers the event stream received from the event sources is precise and certain. Also, application developers neglect the issue of uncertainty in the events. Hence, these applications fail to handle various types of uncertainties in the received events and also affect the complex events derived from them (Diao et al., 2009; Wasserkrug, Gal, Etzion & Turchin, 2012). But the issues related to event refinement still prevail and it is more severe in real time streaming data.
Event Hierarchy depicts the hierarchy of reasoning complex events from low level basic events. The event hierarchy has lower level basic events at bottom and higher level complex events at top.
The Complex Event model is defined by a function y= f(x1,x2,x3,....,xn), where x1,x2,x3,....,xn are parameters (basic events) to function which when applied to the function results into y. The output of y is a complex event as it is associated with n parameters. Accuracy of complex event identification P(y) when the parameters are associated with uncertainty can be obtained with following assumptions:
- 1.
Complex Event model is 100% accurate if the applied parameters are correct. Suppose each parameters x1,x2,x3,....,xn are associated with certainty of u1,u2,u3,...,un then:Maximum possible P(y) = min(u1,u2,u3,...,un) (1) Minimum possible P(y) = 1- (2)
- 2.
Complex Event model is r% accurate if the applied parameters are correct. Suppose each parameters x1,x2,x3,....,xn are associated with certainty of u1,u2,u3,...,un then:Maximum possible P(y) = min(u1,u2,u3,...,un) * r (3) Minimum possible P(y) = 1- * r (4)