A Complex Event Detection Method Based On Macro Forest Transducer

In the era of big data,many applications based on Web generate plethora of detail information as event stream.How to extract valuable information from massive amounts of data and process it becomes a challenge.As a result,some complex event processing platforms were derived.Complex event detection technology is developed to extract valuable information from continuously incoming massive stream data timely and accurately,which is a key component of complex event processing platform.XML(Extensible Markup Language),a format of network data exchange and sharing,is a data model of complex event processing platform.This research on complex event detection methods for XML stream data is fundamentally based on two aspects: first,using a richer complex event description language;second,proposing a more efficient event flow pattern matching method.There are many efficient filtering algorithms for XML stream data,however,their queries are mostly extended on XPath,which are not able to describe the detection of time series relationship in complex event and do not support returning one or more matching sub-results.Complex event detection method,used as a part of complex event processing platform,defined corresponding processing language,but it is complicated in describing complex events.To solve the above problems,firstly,we use the regular tree pattern to describe the complex event pattern,which add regular node to the tree pattern to describe both the structural relationship and time series relationship between the query nodes simultaneously.On this basis,we use a high-performance XML stream data processing model-macro forest transducer as query model of tree pattern,and combine it with Sibling Finite Automaton for regular expression processing.We generate the macro forest transducer and Sibling Finate Automaton based on the regular tree pattern grammar,propose corresponding event stream matching algorithms,implemente a complex event detection system and propose an efficient complex event detection method named CEDMFT.The experimental results proved that it reached 3.3GB/s as its throughput for simple tree pattern queries,and it can reach a throughput of 1.2GB/s for complex queries with time series relationships,which can meet the basic requirements of complex event detection.