Outlier detection of business process based on support vector data description

In order to detect the abnormal status of business process and reduce possible loss, it is necessary to build an outlier detect model. Based on the statistic learning and the support vector classifier theory, a new business processes´ outlier detection model is proposed based on the support vector data description. Firstly, the paper discussed the concept of the business process and the abnormal running status, brought forward the business process´s outlier detection and improvement framework. Then the paper put forward a model for business processes´ abnormal status detection, discussed the problem of parameters selection, detection error and its affect factors. Finally, by simulation method, a set of training data set and testing set is generated, and an experiment is carried for verifying the validity of the model.