A Particle Swarm Optimization-Enhanced Support Vector Regression Model for Accurate Prediction of Concrete Compressive Strength Using Slump Test Data

This paper proposes a hybrid machine learning model combining Radial Basis Function (RBF) kernel-based Support Vector Regression (SVR) with Particle Swarm Optimization (PSO) to predict the compressive strength of concrete using slump test data. Conventional methods rely on labor- and resource-intensive destructive testing, posing challenges for large-scale projects. To address this, SVR models the nonlinear slump-strength relationship, while PSO (swarm size=50, 100 iterations) automates hyperparameter tuning. The SVR-PSO model is benchmarked against Decision Trees, Neural Networks, K-Nearest Neighbors (KNN), and Naïve Bayes, evaluated using R², MAE, MAPE, and RMSE. Results show SVR-PSO achieves and the lowest error rates, reducing prediction costs by up to 40% compared to traditional testing. Limitations include the model’s validation on a specific concrete mix dataset; generalizability to broader formulations requires further study. For reproducibility, code and data will be made publicly available. This work demonstrates how PSO-optimized SVR enables faster, cost-effective strength estimation, supporting data-driven decisions in civil engineering.