This study implements hybrid machine learning models that utilize six commonly employed meta-heuristic algorithms to predict the compressive strength (CS) of manufactured sand concrete (MSC). Six hybrid artificial neural network (ANN) models were created utilizing multiple meta-heuristic algorithms of different groups. A sum of 275 records were used to determine concrete CS of MSC. The hybrid framework, combining ANN and firefly algorithm, i.e., ANN-FF, shows exceptional accuracy in predicting the CS. During the model development stage, the ANN-FF model achieved R² = 0.9536 and RMSE = 0.0498. During testing phase, the values of these indices are R² = 0.9276 and RMSE = 0.0656. The results of the sensitivity analysis demonstrate that the constructed ANN-FF framework effectively estimates the magnitude of the correlation between influential parameters and the CS. The evaluation of outcomes was examined using a variety of tools including Taylor diagram, error matrix, and OBJ criterion. In terms of objective criterion, ANN-FF achieved the best predictive precision. Based on the findings, the constructed ANN-FF can serve as a viable alternative for supporting engineers in civil engineering endeavours. The MATLAB developed ANN-FF model (constructed using eleven distinct influencing parameters) is also attached that can readily be implemented to predict the CS of MSC.

本研究实现了混合机器学习模型，该模型利用六种常用的元启发式算法来预测机制砂混凝土 (MSC) 的抗压强度 (CS)。利用不同组的多种元启发式算法创建了六个混合人工神经网络（ANN）模型。总共 275 条记录用于确定 MSC 的具体 CS。混合框架结合了 ANN 和萤火虫算法，即 ANN-FF，在预测 CS 方面表现出了极高的准确性。在模型开发阶段，ANN-FF模型实现了R ²  = 0.9536和RMSE = 0.0498。在测试阶段，这些指标的值为R ²  = 0.9276 和RMSE = 0.0656。敏感性分析结果表明，所构建的 ANN-FF 框架有效地估计了影响参数与 CS 之间相关性的大小。使用泰勒图、误差矩阵和 OBJ 准则等多种工具对结果进行评估。就客观标准而言，ANN-FF 取得了最佳的预测精度。根据研究结果，构建的 ANN-FF 可以作为支持工程师土木工程工作的可行替代方案。还附带了 MATLAB 开发的 ANN-FF 模型（使用 11 个不同的影响参数构建），可以轻松实现该模型来预测 MSC 的 CS。