Explainable Machine Learning for Intelligent Big Data Analytics in Cloud Computing Environments

Abstract

Machine learning (ML) applied to big data in the cloud now drives decisions about credit, maintenance, fraud, health, and operations. The most accurate models are often the hardest to understand, and at big data scale the problem becomes harder still. Explaining millions of predictions is computationally expensive, explanations must reach audiences with very different needs, and regulators increasingly expect every automated decision to be traceable. This article proposes an architecture for explainable ML in cloud-based big data analytics. It first identifies seven challenges that arise when explainability meets scale. It then sets out a layered architecture that begins with an interpretability-first model selection gate. The architecture provides explanation services using feature attribution and counterfactual methods, and places explanation computation according to latency and cost. It aggregates local explanations into global understanding, delivers explanations tailored to each audience, and routes decisions through configurable workflows. Explanations are stored as audit evidence, and bounded autonomous agents that must themselves explain their actions operate the system. A mapping of audiences to questions, methods, and formats guides explanation design. Using a design-oriented approach grounded in ten studies published between 2016 and 2025, the article illustrates the architecture with an energy utility predicting equipment failures across its grid. It argues that explainability at scale must be designed into the analytics architecture, not added to individual models afterward.