Application of Machine Learning Techniques to Enterprise Model Classification: An Approach and First Experimental Results

Abstract

Nowadays economy forces companies to be adaptive and highly dynamic, what in turn causes a need of often changes in their structures and business processes. Implementation of efficient structures and processes is not possible without proper planning and modelling. Since enterprise modelling is still a highly manual process machine learning is proposed as one of the technologies to assist the modeler and reduce efforts required for model design. Machine learning can be applied to capture regularities in enterprise models and check or suggest relationships and nodes and their types as well as verify or validate the whole models. However, since successful application of machine learning models requires knowledge of the modelling context, the paper is aimed to answer the research question “if it is possible to identify the enterprise model type via machine learning on a limited dataset?”. It was shown that most reasonable approach is to use graph-based representation of enterprise models and application of graph convolution neural networks. The conducted experiments showed that together with natural language processing techniques it is possible to achieve very high accuracy (100% on the dataset used).