Fundamental machine learning algorithms and statistical models applied in strength prediction of geopolymers: a systematic review

The nonlinearity and heterogeneity of geopolymer mix design have urged the research community to supplement the existing experimental design approach with machine learning and empirical regression models to improve the practical strength performance of geopolymers. This systematic review aims to elaborate and consolidate the fundamental machine learning algorithms and statistical models applied in the strength prediction of geopolymers. This review specifically delves into the statistical linear/nonlinear optimization algorithms, supervised machine learning algorithms, and model performance statistical metrics. The PRISMA and Scopus databases were used for bibliometric data extraction. The search strings devised to carry out the review were “geopolymer” OR “alkali-activated materials” AND “machine learning” OR “statistical modeling”. This review observed that neural networks, random forest, support vector machines, and Gaussian process regression give better strength prediction performances with R² values > 0.9 and RMSE values < 3. The choice of activation function influenced the training performance of the algorithm and defined the prediction output accuracy. Hyperparameter tuning and Shapley additive explanations showed that input features with a greater effect on compressive strength were curing conditions and silicate modulus. This review promotes the consolidation of conventional experimental mix design approaches with machine learning techniques in solving geopolymer mix design and strength-related problems to give greater confidence to engineers and researchers in the applicability and versatility of these models to real-life practical scenarios saving time and minimizing costs.