Research On Predictive Model Of Complex Event For The Hub Airport

Voluminous of event logs have accumulated in the daily operation of hub airport. With these event logs, the business intelligence technology like process mining can be used to analyze the airport events. By aggregating the low-level events, the airport events are reorganized as the complex events, which can be used to construct a predictive model. Interesting results will be generated and a predictive solution for airport operation can be built. Therefore it is important to research on predictive model of complex event for the hub airport.To analyze the airport event logs by quantifying the attribute of events, this paper build a bi-type information network to describe the relationship between attributes of event like organizational entities and cases of instaniated process. A method to sort the quantified importance of organizational entities organized under the trace clusters is proposed to prove the effectiveness of the network, which is based on the characteristic that similarity between the traces of cases exists. Experiments show that this method can express the actual importance of the attributes more precisely.The airport event logs are on a detailed level of abstraction. An approach to overcome this is to group low-level events into clusters, which represent the execution of a higher-level activity in the process model. Therefore, this paper presents a new activity mining method which is based on RankClus algorithm to generate activity clusters integrated with ranking. On this basis, the complex events can be extracted as the high-level activity clusters. The experiment results show that this activity-clustered model is significantly less complex which shares a similar level of conformance with the meta model.After extracting the complex events, the prediction of complex events can be transformed into a curve fitting problem of the attributes' importance. A feed-forward neural network trained by the Levenberg-Marquardt algorithm can be used to solve this problem. Compared with the model trained by the Scaled Conjugate Gradient algorithm, this model is more precise, meanwhile it has a better performance index.