Sequential learning for case-based pattern recognition in complex event domains

Large and distributed environments usually generate huge amounts of information. The easiest way to deal with this information in an uncoupled and asynchronous way is using event oriented approaches. These systems are usually implemented to react to the generated information. This paper presents a new track to add to these architectures a mechanism to discover behaviors combined with a reasoning method that predicts the next most probable event. Consequently, the work focuses in two fields: sequence pattern mining and case-based reasoning. The former aims to compress the original large amount of event data by discovering frequent behaviors in the form of sequence patterns. The latter is used to recognize the behaviors and forecast future predictions based on the learnt patterns. The methodology has been tested using real data from a public bike hiring