How Machine Learning Improves Enterprise Architecture Models?

Oct 19, 2022
Written by
Maarten Steen
Maarten Steen

How Machine Learning Improves Enterprise Architecture Models?

Machine Learning (ML) is a subfield of artificial intelligence aiming to teach computers to think the same way humans do. Data is used to train an algorithm that identifies and discovers patterns and makes predictions for new cases. Using ML is not uncommon. Examples include financial institutions using ML to detect potential cases of fraud. Online retailers use ML to recommend relevant products to their customers. ML is also used in logistics to predict demand, plan capacity, and many more.
Bizzdesign is using ML to improve the user experience for enterprise architecture designers and users of our Bizzdesign Horizzon platform. We’ve created a beta version of our ML-based feature called the Relation Recommender for ArchiMate, which is now available for customer testing.

What is Relation Recommender for ArchiMate? How can you use it?

ArchiMate is a powerful, graphical modeling language for enterprise architecture, standardized by the Open Group and the most widely used language on our platform. ArchiMate provides concepts for capturing an enterprise’s strategy, business, application and technology designs. To express the relationships between objects, ArchiMate also provides 11 different relation types, which are listed in the table below.

The richness of the ArchiMate language enables accurate modeling, but it could also be challenging for beginners and occasional modelers. For example: How do you decide on the most appropriate relation between two objects in your model? This decision requires training and experience.

Bizzdesign Horizzon already helps users by blocking the creation of relations that aren’t allowed according to the ArchiMate metamodel and by only suggesting ‘legal’ relations when using the Smart Connector. However, there are still many options to choose from. The right choice depends on your organization’s modeling guidelines and conventions and your intention as a modeler.

Smart Connector only proposes relations allowed on the current view type and between the selected source and target object types 

We’re now introducing the ML-powered Smart Connector. When you switch on the new Relation Recommender, the Smart Connector becomes even smarter. It picks the most appropriate relation type for you depending on what it learned from the content already in your model package. You just connect the objects you want to relate to, and the Smart Connector will add the relation that matches your organization’s best practice for the given context.

Now that you don’t have to think about the relation type anymore, you can model much faster. Also, your less experienced colleagues can add relations to the models without worrying about using the right relation type, thereby increasing their quality and conformance to your modeling guidelines. And, if you disagree with the selection made by the algorithm, you can always change the relation type afterward using the context menu. Additionally, suppose the algorithm accidentally picks a relation type that isn’t allowed according to the metamodel, for example, because it has not seen enough examples during its training. In this case, the Smart Connector falls back to its default behavior and asks you to choose.

Want to try it out for yourself? Get in touch with Bizzdesign Support to enable the Relation Recommender for ArchiMate in your instance of the Bizzdesign Horizzon platform.

We would like to know how you experience this machine learning-supported way of modeling. Is the Relation Recommender helping you to model better and faster? Would this feature help your less experienced colleagues to model correctly in ArchiMate? Let us know via the Bizzdesign Community.