or it can return as an input to the CEP engine.
The event patter’s role is to specify how the incoming
events should be processed in order to extract relevant
information. The language used to define these patterns
should have the ability of specifying complex relationships
among events flowing into the CEP engine.
The typical approach in defining patterns of events is to
manually specify them. This is done either by domain
experts, capable of providing the definition of event patters
or by using other tools externally of the CEP systems in
order to discover these patterns and then encode them in the
event processing language (EPL). However, we see the
integration of machine learning algorithms with the CEP
system, as a solution for direct support in definition of
event patterns. Although massive amount of research has
been conducted in the areas such as pattern recognition and
multisensor data fusion, the systems developed for many of
the CEP applications do not provide a seamless integration
with such techniques, but rather consider the human
component responsible for defining the complex events
patters that should be monitored and detected. Therefore,
an important improvement for applying machine learning
algorithms in event-based application is to develop a
framework that would allow easy integration of existing
algorithms with event processing techniques.
A first step in achieving such integration is choosing a
scenario for running experiments. One example is the smart
cities scenarios, as there can identified many data sources
and use cases for data mining and CEP. In this paper we are
proposing such a scenario, identify the data sources and run
preliminary experiments for analysing the data. Future steps
are discussed in the direction of using CEP engines with the
patterns discovered and defining a framework for an easier
integration of data mining and CEP.
The rest of the paper is structured as follows: Section 2
describes the smart cities scenario and introduces one use
case considering the city of London. Data integration and
preprocessing is presented in Section 3, while Section 4
discusses the result of data mining. Finally we conclude the
paper and identify future directions.