Comparative ML Machine Learning Models for Concrete Slump Prediction: A Performance Analysis

This research presents a comprehensive comparison of Naive Bayes, Decision Trees (J48), and Multilayer Perceptron (MLP) Neural Networks in accurately predicting the slump of fresh concrete. Utilizing a dataset of 103 samples, we evaluate the performance of these models, demonstrating the Decision Tree's ability to identify crucial predictors and the MLP's exceptional predictive accuracy. Our findings underscore the MLP's superiority in classifying concrete slump based on various metrics, including accuracy, error rates, and ROC/PRC curve areas. Furthermore, we investigate into the correlation between concrete ingredients and slump, emphasizing the significant impact of water content. By combining domain knowledge with machine learning, our research provides valuable insights for optimizing concrete mix design and improving construction efficiency.