Data-Mining Dynamical Systems: Automated Symbolic System Identification for Exploratory Analysis

This paper describes a new algorithm for automatically reverse-engineering symbolic analytical models of dynamical systems directly from experimental observations, for the purpose of modeling, control and exploratory analysis. The new algorithm builds on genetic programming techniques used in symbolic regression to infer differential equations from time series data. We introduce the core algorithm for building coherent mathematical models efficiently and then describe its application to system identification. The method is demonstrated on a number of nonlinear mechanical and biological systems.