Performance analysis of linear and non-linear machine learning models for forecasting compressive strength of concrete

Abstract

The compressive strength of concrete is a critical parameter that affects the safety, durability, and performance of structures. This is a key factor in the design, construction quality control, and long-term functionality of concrete elements in various engineering applications. This is a fundamental indicator of the ability of the concrete to withstand axial loads or pressures. Understanding the compressive strength is essential for ensuring the safety of structures. This helps engineers design concrete elements with sufficient load-bearing capacity to withstand the expected loads and maintain structural stability. The objective of this study is to assess the effectiveness of linear and non-linear machine learning algorithms in accurately predicting the compressive strength of concrete based on varying quantities of constituent components. To address outliers in the data, we utilized an iterative implicit technique during the data preprocessing stage. Our implementation included Linear models such as Linear regression and polynomial regression, Non-linear models such as Decision Tree and Random Forest, and Boosting models such as Gradient Boosting and Extreme Gradient Boosting. The accuracy scores and overfitting of the models were compared. The models were evaluated using the R2 scores for both the training and test sets. Among the three categories of models tested, boosting models demonstrated higher accuracy than the other models, while also exhibiting limited overfitting.

Keywords

• Linear model
• non-linear model
• regression
• R2 score
• tree models
• iterative imputer