Functions Provided by MATLAB Neural Network Toolbox

Using functions provided by MATLAB's Neural Network Toolbox, we can implement both BP (Backpropagation) and RBF (Radial Basis Function) neural networks to establish mathematical models between near-infrared spectra of gasoline samples and their corresponding octane values. Key functions such as `feedforwardnet` for BP networks and `newrb` for RBF networks can be utilized with appropriate parameter tuning. The performance evaluation of these models involves metrics like mean squared error (MSE) and regression analysis, implemented through functions like `perform` and `regression`. These validated models enable accurate prediction of octane values for unknown samples, playing a crucial role in gasoline quality control and inspection processes. Furthermore, the neural network modeling approach facilitates exploration of additional correlations and causal relationships, expanding our understanding of gasoline sample characteristics through feature importance analysis and sensitivity testing.