UCLA

In this study, I present a comprehensive study that addresses the complex challenge of predicting concrete strength, leveraging the power of advanced machine learning techniques. Recognizing the limitations of traditional prediction models, I have introduced innovative methodologies to enhance accuracy and interpretability in this crucial aspect of construction.

Central to my approach is the development of the Ensemble-Based Outlier Detection (EBOD) algorithm. Recognizing the detrimental impact of noisy data on model performance, I designed EBOD to integrate multiple detection algorithms, thereby significantly reducing the bias associated with single-algorithm methods. This innovation ensures that the datasets used for model training and analysis are of the highest quality, laying a solid foundation for more accurate predictive modeling.

Moving forward, I explored the capabilities of Gaussian Process Regression (GPR) in predicting concrete strength. My work with GPR is not just about prediction; it's about understanding the intricacies of the data. I optimized the GPR model to not only forecast concrete strength with remarkable accuracy but also to quantify the uncertainties associated with these predictions. This dual capability of the GPR model enriches the interpretability of the results, providing deeper insights that are invaluable for material engineering and construction management.

In my pursuit of transparency and interpretability in predictive modeling, I introduced symbolic regression into the study. I recognized the need for models that not only predict but also explain. Symbolic regression offered a solution, enabling me to construct interpretable models that shed light on the underlying physical phenomena governing concrete strength. To enhance the predictive power of these models, I incorporated advanced data augmentation techniques, such as the Synthetic Minority Over-sampling Technique (SMOTE), pushing the boundaries of prediction and understanding in unexplored domains.

A pivotal aspect of my study involved a meticulous analysis of the balance between data volume and the precision of machine learning models. I undertook a comprehensive evaluation of a vast dataset, assessing the performance of various algorithms in predicting concrete strength. This rigorous analysis highlights my commitment to not only advancing the accuracy of predictive models but also to understanding the practical challenges and limitations of employing machine learning in the field of concrete strength prediction.

Through the development of innovative algorithms, the application of advanced machine learning techniques, and a thorough analysis of extensive datasets, I aim to revolutionize the way we predict, understand, and apply concrete strength models in industrial applications, setting new benchmarks for accuracy and interpretability.