PSI - Issue 68

Hussam Safieh et al. / Procedia Structural Integrity 68 (2025) 245 – 251 H. Safieh et al./ Structural Integrity Procedia 00 (2019) 000–000

249

5

3. Results and Discussion

Table 1: Model results comparison

Model

MSE

MAE

RMSE 11.9404 12.2161 6.5392 7.0579 11.9347 11.6477 11.5638 11.8965 12.4818 11.9404

R²

Linear Regression

142.5742 149.2324 42.7610 49.8142 142.4382 135.6690 133.7217 141.5263 155.7942 142.5742

9.8903 9.8659 4.3891 5.3921 9.8885 9.9554 9.9141 9.8762 10.0113 9.8903

0.2974 0.2646 0.7893 0.7545

Linear Support Vector Machine

Random Forest

k-Nearest Neighbors Ridge Regression Lasso Regression ElasticNet Regression

0.298

0.3314

0.341

Bayesian Ridge Regression

0.3025 0.2322 0.2974

Huber Regression Lars Regression

In this study, several machine learning models were trained and tested to predict the compressive strength of concrete with varying fly ash percentages. The models compared include Linear Regression, Linear Support Vector Machine (SVM), k-Nearest Neighbors (KNN), Ridge Regression, Lasso Regression, ElasticNet Regression, Bayesian Ridge Regression, Huber Regression, and Random Forest (RF). As summarized in Table 1, the Random Forest model demonstrated the highest accuracy with an R² of 0.7893, significantly outperforming all other models. The Random Forest model's accuracy can be attributed to its ensemble learning approach, which mitigates overfitting by aggregating the results of multiple decision trees, thus capturing complex nonlinear relationships in the dataset more effectively. This is particularly evident in the sharp reduction in Mean Squared Error (MSE) and Root Mean Squared Error (RMSE), as well as a lower Mean Absolute Error (MAE), compared to other models. The RF model is known for its robustness in handling high-dimensional data, its ability to capture both linear and nonlinear patterns, and its effectiveness in reducing overfitting, which was particularly crucial in this study due to the variability introduced by the varying fly ash content. Moreover, the feature importance analysis conducted through Random Forest revealed that the fly ash percentage (FA%) and curing period were the most significant factors influencing the compressive strength prediction. This aligns with previous studies, which have emphasized that higher FA% enhances the concrete's long-term strength, especially in combination with prolonged curing periods. The scatter plot in Figure 2, which compares the training vs. test values, shows that Random Forest predictions align closely with the actual compressive strength values, further reinforcing its suitability for this type of data.