Classification of steel samples by laser-induced breakdown spectroscopy and random forest

Laser-induced breakdown spectroscopy (LIBS) integrated with random forest (RF) was developed and applied for the identification and discrimination of nine steel grades. The classification and recognition of the steel grade was completed by investigating their physical and chemical properties. Two parameters of the RF were optimized by out-of-bag (OOB) estimation. The generation ability of RF model was evaluated by OOB estimation and 5-fold cross-validation (CV). Compared with the partial least squares discriminant analysis (PLS-DA) and support vector machines (SVM), the classification of steel samples based on RF model shows a better predictive performance. It has been confirmed that RF method is promising for automatic real-time, fast, reliable, and robust measurements; and the system can be integrated into portable form for non-specialist users.