Using Machine Learning Techniques to Improve Information Systems Development Methods

摘要

We have presented a learning method for process model improvement. The originality of our work lies in the combination of process meta-modeling with machine learning techniques: the learning method is modeled as an instance of the decision-oriented process meta-model and uses induction techniques to guide the learner and partly automate learning.rnOur work is described more thoroughly in [PRAT95b]. We have started implementation in the MENTOR prototype. In the immediate future, we will concentrate on further formalization and experimentation of the learning method.rnAlthough our primary focus is process model improvement, the use of induction techniques is also relevant for trace-based process model definition.rnIn addition to induction, we believe that case-based/analogical learning techniques could be used in our repository-based framework: case-based reasoning techniques could be applied to develop new applications (solve new problems) by reusing process traces (solutions), and analogical reasoning techniques could be adapted to define new process models by analogy with existing ones.