Validation of parameter importance by regression

This article employs regression techniques in order to gauge the relative importance of causal factors affecting the emergent behaviour of a hybrid dynamical system simulating human behaviour. Extending our previous work, in which the relative importance of causal factors was determined on the basis of classification into two discrete clusters, the target characteristic attribute of the simulation is now represented by a non-negative real number. This enables us to represent the dominant simulation state with higher sensitivity and apply various regression methods: Neural network - Multi-layer perceptron regressor, Radial basis function regressor and M5P regression tree.