Predictive analysis of foam concrete compressive strength: a comparative study of OLS and SVR with K-fold validation

Abstract

Foam concrete (FC) is a building material created by blending foam along with cement mortar. FC being a light weight material aids in achieving low density and high durability which can be used in multiple building scenarios. This study explores the enhancement of FC by incorporating coarse aggregate (in ratio) and varying proportions of silica fume (as a percentage) with a constant 3% foam volume. The aim is to study the correlation between the variables using statistical analysis such as descriptive statistics, correlation matrix, and one-way ANOVA. From the results of statistical analysis, Compressive Strength (CS) of FC is predicted using two different regression techniques: Ordinary Least Square Regression (OLS) and Support Vector Regression (SVR). These techniques are employed to forecast the compressive strength of FC. The dataset consists of seven independent variables and one dependent variable. From the statistical analysis it is observed that there is high correlation between dependent and independent variables. OLS regression demonstrates superior predictive capabilities compared to SVR. In addition to regression techniques, a validation study known as K-fold cross-validation assesses the model's predictive performance and generalization reliability, highlighting their potential for optimizing high-strength FC formulations.