业务流程中一种个性化的任务完成时间预测方法

Approach for individual task completion time prediction in business processes

为实现业务流程中人力资源的有效管理与调度,提出一种个性化任务完成时间的预测方法。该方法首先分析了业务流程的历史日志,对资源的状态和能力进行了建模与评估;然后,采用支持向量机回归的方法,发挥其"柔性边界"和核函数的特点,对日志数据进行预处理、模型训练及参数优化,再结合资源当前的状态和历史能力对资源完成任务的时间进行实时预测。同时考虑了稀疏数据集问题的处理。实验表明,在日志记录充分时,尤其是在完成时间波动较大的情况下,该方法的预测结果能达到较高的准确率,可为资源管理、任务分派等过程提供实时的决策依据。

To manage and schedule human resources effectively in business processes,an approach for individual task completion time prediction was proposed.In this method,the status and capability of resources were modeled and evaluated based on the analysis of historical logs.The algorithm of Support Vector Regression(SVR)was adopted,and its advantages of“soft margin”and various kernel functions were developed to realize the real-time prediction for task completion time through the process of data integration,data standardization,model training,parameter optimization and so on,which had considered the current status and the capacity of specific resources.The problems of dealing with sparse data sets were also taken into consideration.The experiment result revealed that the prediction reached a higher accuracy with sufficient logs according to the approach,which would provide real time decisions for the process of resource management and task assignment,especially in the case of high fluctuation of completion time.