The Logical Glue platform allows you to be build classification or regression models using Type-2 Fuzzy Logic. In order to create these models, a vast set of IF…THEN type rules is automatically created from your data (such as IF TimeOnSite = high AND LinksClicked = high THEN ProductPurchase = TRUE). All of those possible rules are then honed down using machine learning to find those that are most predictive of the use case and that work best together. A final set of rules, ordered and weighted by their predictive power, constitutes the model.
To make a prediction for a particular instance (for example a loan application), individual rules are “fired” according to the various input characteristics and contribute a certain weight to the overall score. This means that all predictions are completely transparent and easy to understand by a non-technical user.
Not only are these Type-2 Fuzzy Logic models powerful in their own right, outperforming traditional statistical techniques such as logistic regression, but they can be combined with other machine learning techniques within the platform (such as neural networks) to maximise both model accuracy and interpretability.