Analysis of Strength Prediction Models using Machine Learning for Mongolian Fly Ash Concrete

Abstract

This paper investigates the application of machine learning (ML) methods for predicting the compressive strength of Mongolian fly ash concrete. Four predictive models—Multiple Linear Regression (MLR), Ridge Regression, Lasso Regression, and Decision Tree Regression—were developed and compared using both experimental data and benchmark datasets. The decision tree model exhibited the highest predictive accuracy, achieving an R² of 0.95 and significantly lower RMSE and MAPE compared to linear regression-based models. Results highlight that Mongolian fly ash concretes demonstrate distinct strength development behavior compared to global trends, underscoring the importance of region-specific prediction models for safe and economical structural design.