The global construction industry is at a pivotal juncture, pressured to mitigate its substantial environmental footprint while meeting escalating infrastructure demands. A promising pathway is the incorporation of industrial and agricultural waste by-products as supplementary cementitious materials (SCMs) or aggregates in concrete and asphalt. While hundreds of individual studies have investigated these materials, the literature remains fragmented, often yielding contradictory conclusions regarding optimal replacement levels and performance outcomes. This paper proposes a novel paradigm: a large-scale meta-analysis and knowledge synthesis framework that leverages Machine Learning (ML) algorithms and Information Value (IV) models to unify these disparate findings. Instead of conducting new experiments, this review synthesizes data from existing literature, including 26 exemplar studies, to identify global trends, hidden correlations, and quantitatively rank the information value of different waste materials. We explore the application of advanced ML techniques including Frequency Ratio (FR), Logistic Regression (LR), Artificial Neural Networks (ANN), and Weight of Evidence (WOE) traditionally used in geospatial analysis (e.g., landslide susceptibility mapping) to the domain of material informatics. The core objective is to transition from qualitative, experience-based material selection to a quantitative, data-driven decision-support system. This synthesis demonstrates that ML-powered meta-analysis can pinpoint optimal waste material incorporation ratios, predict long-term performance, and ultimately accelerate the adoption of sustainable, high-performance construction materials by providing a robust, evidence-based foundation for engineers and researchers.