Hybrid Response Surface Methodology-Support Vector Machine modeling approach enhanced the efficiency of optimization for sansevieria propagation
摘要
The plant tissue culture process is significantly influenced by various endogenous and environmental parameters that jointly regulate plantlet development. This study employed a hybrid modeling approach combining Response Surface Methodology (RSM) and Support Vector Machine (SVM) to predict and optimize the effects of plant growth regulators on the propagation of Sansevieria trifasciata Prain, a species characterized by slow growth and high market demand. An experimental design generated by RSM provided the data for training the SVM algorithm, which was used to determine importance coefficients. The linear kernel SVM achieved an accuracy of 0.99 in revealing interactions within the input data. The importance coefficients indicated that the callus state negatively affected plantlet production, whereas root formation had the strongest positive influence. An optimal hormonal combination of 0.2 mg L−1 2,4-D, 0.18 mg L−1 NAA, and 1 mg L−1 KN (desirability = 1.00) was identified and validated through both experimental results and SVM-RSM predictive analyses, confirming the model robustness and the method reliability. The best model was supported by R2 = 83.32, R2adj = 79.03, R2pre = 72.37, and adequacy precision of 15.70. These findings represent a significant advancement, reducing the production time for sansevieria plantlets to approximately 3.2 months, compared to 5–6 months using traditional methods. This study highlights the effectiveness of the hybrid RSM-SVM approach in plant tissue culture, which leverages the complementary strengths of both methods for optimizing factor levels (i.e., RSM) and for robust binary classification of outcomes (i.e., SVM). It exemplifies how artificial intelligence (AI)-driven optimization can enhance the productivity and efficiency of plant tissue culture protocols.