Simulation Testing of SVR Regression Model Implementation

In this paper, we utilize MATLAB and functions from the libsvm package to implement simulation testing of Support Vector Regression (SVR) models, applying it to predict concrete compressive strength. Our implementation includes key steps such as data normalization using z-score standardization, kernel function selection (RBF kernel), and parameter tuning via the svmtrain function with epsilon-SVR type specification. We conducted comprehensive experiments and analysis to validate the algorithm's effectiveness and accuracy, employing performance metrics like Mean Squared Error (MSE) and R-squared values. Additionally, we explored the algorithm's potential applications in other domains to demonstrate its broad applicability. Our research provides a novel methodology and tool for strength prediction in concrete engineering, offering valuable references for related research and practical applications. The code structure incorporates feature scaling, model training with optimal parameters obtained through cross-validation, and prediction visualization using MATLAB's plotting capabilities.