Incremental outcome-oriented predictive process monitoring based on XGBoost

With the improvement of industrial manufacturing business processes, monitoring technology aimed at predicting the results of execution is necessary. The technique builds prediction models based on historical execution to predict the results of the processes being executed. However, existing studies assume that the process execution behavior remains the same, but the process often changes during the operation (the process execution drift) in practical application, so the prediction model needs to adapt to this drift. In response to this situation, inspired by the i-dea of online learning, a predictive process monitoring technology was proposed based on XGBoost incremental implementation targeting process execution outcomes, and a large number of experiments on real data sets and synthetic data sets were conducted respectively. The experimental results showed that the incremental learning technology based on XGBoost could well provide an effective solution for predictive process monitoring in real scenarios of industrial manufacturing. © 2024 CIMS. All rights reserved.