Predicting the Tensile Strength of Plant Leaves Based on GA-SVM

Most plant material mechanics studies focus on the independent investigation of single properties, whereas systematic research on the coupling relationships among multidimensional performance indicators of plant leaves remains relatively scarce. This study employs a comprehensive testing instrument to characterize the physical and chemical properties of plant leaves, tensile tests to assess tensile strength, and high-speed cameras to capture crack propagation during leaf fracture. Subsequently, correlation analysis is performed on the characteristic indicators, followed by dimensionality reduction using principal component analysis (PCA). A genetic algorithm (GA) is then applied to optimize the structural parameters of the support vector machine (SVM), establishing a GA-SVM-based predictive model for the tensile strength of plant leaves. A comparative analysis with other predictive algorithms demonstrates that the GA-SVM model achieves the lowest prediction error and highest accuracy, with mean absolute error and root mean squared error values of 0.0774 and 0.0745, respectively. Among the five models evaluated, the prediction accuracy ranks as follows: GA-SVM, SVM, decision tree, random forest, and back-propagation neural network. These findings fully validate the effectiveness and superiority of the GA-SVM model in predicting the tensile strength of plant leaves, providing a novel methodology for accurately predicting mechanical properties of plant materials.