An approach for determining relative input parameter importance and significance in artificial neural networks

Artificial neural network (ANN) models are powerful statistical tools which are increasingly used in modeling complex ecological systems. For interpretation of ANN models, a means of evaluating how systemic parameters contribute to model output is essential. Developing a robust, systematic method for interpreting ANN models is the subject of much current research. We propose a method using sequential randomization of input parameters to determine the relative proportion to which each input variable contributes to the predictive ability of the ANN model (termed the holdback input randomization method or HIPR method). Validity of the method was assessed using a simulated data set in which the relationship between input parameters and output parameters were completely known. Simulated data sets were generated with known linear, nonlinear, and collinear relationships. The HIPR method was performed repetitively on ANN models trained on these data sets. The method was successful in predicting rank order of importance on all data sets, performing as well as or better than the recently proposed connectivity weight method. One main advantage of using this method relative to others is that results can be obtained without making assumptions regarding the architecture of the ANN model used. These results also serve to illustrate the consistency and information content of ANN models in general, and highlight their potential use in exploring ecological relationships. The HIPR method is a robust, simple, general procedure for interpreting complex ecological systems as captured by ANN models.