Green Data Mining Algorithms for Energy-Efficient Large-Scale Knowledge Discovery: A Comprehensive Review

Abstract

The exponential growth of big data across enterprise, industrial, and scientific domains has made large-scale knowledge discovery a cornerstone of modern intelligent decision-making. However, extracting actionable patterns from petabyte-scale datasets imposes an unprecedented dynamic energy burden on high-performance compute clusters and hyper-scale data centers. Traditional data mining algorithms—including iterative clustering, deep frequent itemset mining, and multi-dimensional graph processing—were engineered primarily for runtime speed and algorithmic precision, frequently ignoring power dissipation profiles, memory dynamic dynamic allocation efficiency, and dynamic CPU/GPU dynamic frequency scaling. This review delivers a rigorous, state-of-the-art synthesis of Green Data Mining (GDM) algorithms designed specifically for energy-efficient, sustainable knowledge discovery. We systematically evaluate fundamental algorithmic paradigms, including energy-aware approximation techniques, green distributed MapReduce/Spark job schedulers, energy-constrained pattern pruning, and dynamic voltage and frequency scaling (DVFS)-aware execution engines. Furthermore, we survey computational frameworks that integrate dynamic energy-delay product (EDP) metrics directly into objective optimization functions. Experimental, numerical, and comparative benchmarks demonstrate that green data mining frameworks can yield up to 60% reductions in carbon footprint and energy consumption while sacrificing under 2% knowledge extraction accuracy. Finally, critical gaps in energy-aware multi-tenant resource contention, neuromorphic hardware integration, and green edge computing are highlighted alongside future actionable research trajectories. KEYWORDS: Green Data Mining; Sustainable Computing; KnowledgeDiscovery; Energy-Aware Schedulers; Energy-Delay Product; Big DataAnalytics.