This thesis explores the application of Artificial Neural Networks (ANNs) for predicting the compressive strength of concrete, a critical parameter in construction engineering. Given the complexity of concrete's composition and the various factors influencing its strength, traditional methods for predicting compressive strength often fall short. This study aims to leverage the capabilities of three different ANN architectures: Feedforward Neural Networks (FFNN), Recurrent Neural Networks (RNN), and Radial Basis Function Neural Networks (RBFNN) to enhance prediction accuracy. We collected a dataset of 782 samples from nine different research papers, each containing 12 input features relevant to concrete mix design and one output feature representing the compressive strength. The dataset was normalized using z-score normalization and cleaned by removing outliers. We split the data into training, validation, and test sets to ensure robust model evaluation. Our FFNN model achieved a Mean Squared Error (MSE) of 0.001522, a Root Mean Squared Error (RMSE) of 0.038995, a Mean Absolute Error (MAE) of 0.029068, a Relative Squared Error (RSE) of 0.541426, a correlation (R) of 0.736775, and an R-squared (R^2) of 0.542837. The RBFNN model performed better, with an MSE of 0.000727, an RMSE of 0.027216, an MAE of 0.016682, an RSE of 0.0269, an R of 0.91587, and an R^2 of 0.838721. The RNN model yielded similar results to the FFNN, with an MSE of 0.002071, an RMSE of 0.045506, an MAE of 0.033741, an RSE of 0.607956, an R of 0.725412, and an R^2 of 0.52622. The findings demonstrate that the RBFNN outperforms both the FFNN and RNN in predicting concrete compressive strength, highlighting the potential of ANNs in enhancing predictive accuracy in civil engineering applications. Our research contributes to the limited body of work in this area and offers a foundation for future studies to build upon. This work underscores the versatility and effectiveness of ANNs in solving complex, nonlinear problems across various domains.