Mohd Hafeez Osman – An Analysis of Machine Learning Algorithms for Condensing Reverse Engineered Class Diagrams

The aim of this work is to automatically simplify UML diagrams. Why would you want that? Well, extracted diagram can often be too complex to quickly grasp, especially for newcomers. For this, they use a supervised machine learning approach, in which the design is used as a source to learn from. Their basic strategy works as follows:


Their evaluation on the following 9 open source projects


showed that Random Forest and k-Nearest Neighbor provided the best results. For the case studies, the Random Forest method scores an AUC above 0.64 and the average AUCs for every prediction set is 0.74.

No pre-print available unfortunately.

This post was visited 191 times.