Towards ant colony optimization of neuro-fuzzy interval rules

Neuro-fuzzy rules can be used in their fuzzy form and in an interval form that is a cut of the corresponding membership function. Such interval rules can be derived whenever a precise interval rule is useful in the application area. An example where interval rules can be applied is the area of virtual screening in chemistry. Current research focusses on finding novel drugs. Nowadays, preselection of potential molecules can be done by computational analysis of molecular properties. Usually, a high-dimensional descriptor vector represents the molecular properties for one molecule. With a well-established neuro-fuzzy system that is capable of selecting important features, we describe the process of interval rule generation within the application. Since the neuro-fuzzy interval rules need not to be optimal, the idea of ant colony optimization is adapted for solving the interval rule optimization problem. The results demonstrate the capability of interval rule optimization by an ant colony but also its dependency on the number of ants.