Evaluation Method based on Random Forests for Empirical Algorithms of Penetration Effects

Experimental analysis and derived empirical algorithms are the most commonly methods for the analysis of damage effects. Quantity and quality of experimental data and the derivation theory resulting in a limited scope of application of empirical algorithms. The existing literature descriptions of the scope of application of empirical algorithms are often qualitative, making it difficult to perform destruction effect calculations quickly select and effectively in practical engineering application. In this paper, an empirical algorithm evaluation method is proposed for concrete penetration effects algorithms based on random forest and experimental data. Multiple base-classifiers are constructed with multiple sampling, and integrated to reduce the dependence on data sets. The experiment results show that the proposed method can evaluate the empirical algorithms more accurately than directly compute average error on experimental data.