PSI - Issue 70

Indu Sharma et al. / Procedia Structural Integrity 70 (2025) 380–385

381

1. Introduction Pre Ordinary Portland cement is the most commonly utilized concrete cement, which is energy-intensive and produces CO 2 emissions Chen et al. (2022) and Khan et al. (2022). OPC manufacturing is responsible for the emission of 4 billion tons of CO2 each year, accounting for 5 – 7% of the global CO2 output Biricik et al. (2021) and Mahmoodet al. (2021). In light of environmental and climate change issues, numerous projects have reduced the production and use of OPC Farooq et al. (2021). The use of additional cementitious materials and the recycling of waste contribute to a reduction in building OPC Tariq et al (2022). The percentage for SCM are relatively modest when compared to OPC Miller (2018). Fly ash exhibits pozzolanic properties during the hydration of ordinary Portland cement, yet it does not influence early strength. Fly ash inhibits hydration and setting Wu et al. (2021), limiting its application. Eco-friendly alkali-activated cementitious binders are frequently used as substitutes for OPC Pacheco et al. (2008), high-energy calcination that OPC clinker does. Geopolymers consist of three-dimensional spatial complex polymeric alum inosilicate cementing components that are activated using NaOH or Na2SiO3 derived from industrial wastes fly ash Algaifi et al. (2021). The chemical composition of geopolymers enhances mechanical performance and durability. Using waste as a binder renders these methods more environment friendly compared to OPC-based mixtures Okoye (2017). This study investigates GPC CS through the lens of ML. We employed R2, k-fold, statistical tests, and predicted error divergence from experimental data to evaluate SVM represent ensemble machine learning methods, while SVM is categorized as a single model Yeh and Lien (2009). This exceptional study forecasts GPC CS by employing both single and ensemble machine learning algorithms. Conducting experimental research necessitates financial resources, time investment, and effort for the collection of ingredients, as well as for casting, curing, and testing processes. Innovative methods such as ML will help the construction sector tackle these challenges. Due to the influence of precursors, activators, aggregates, and similar factors on GPC's CS, experimental approaches present challenges. Simple ML algorithms might reveal the collective impact of research components. Given that multiple experiments have demonstrated GPC's for CS, it is possible that ML algorithms could utilize book data. Data trains machine learning algorithms to forecast material properties. Earlier research assessed the mechanical strength of GPC by employing machine learning algorithms with fewer input parameters and samples. Chaabene and Flah (2020) utilized machine learning to predict GPC CS, employing three input parameters and a dataset comprising data points. A different investigation employed four input parameters along with data samples Dao et al. (2019). Many researchers (Nikhil et al. 2023, Malik et al. 2025) have used ML models in solving civil engineering problems. This research forecasted GPC CS by utilizing nine input parameters, data samples, and various ML algorithms. This research explores the best machine learning approach for forecasting GPC for CS and study the influence of various factors on GPC strengths. Comparative analysis of novel ML approaches aims to enhance the precision of GPC CS predictions in future studies. These methods enable researchers and builders to conserve time, reduce costs, and minimize experimentation. 2.Methods 2.1. Data Retrieval and Analysis Effective SML requires various input variables. The CS of GPC was established based on academic literature. Data from the literature was selected at random to ensure impartiality. This study applied algorithms to data points based on computer science, while other investigations focused on the characteristics of GPC. CS is generated through algorithms that incorporate GGBS, fly ash, NaOH molarity fine aggregate, water ratio, and 10/20 mm gravel. The model output is significantly influenced by inputs and datasets Han et al. (2019). The investigation of the Supplementary Materials employed ML algorithms on a dataset comprising 371 data points. To gather data, mix proportions and the desired outcome were utilized, as models required consistent input parameters for each