Big Data Analytics for Intelligent Decision Support Systems

Abstract

Intelligent decision support systems (IDSS) that leverage big data analytics are transforming organisational decision-making across enterprise,
healthcare, financial, and public sector domains by integrating real-time streaming analytics, machine learning inference, and natural language
interfaces into actionable decision workflows. Despite widespread deployment of individual big data technologies, unified frameworks that coherently
address the full IDSS pipeline — from large-scale heterogeneous data ingestion and real-time stream processing through ML-driven insight
generation to explainable recommendation delivery — remain architecturally fragmented. This study presents DataMind, a unified big data analytics
framework for IDSS integrating Apache Kafka-based stream processing, a distributed feature store, an AutoML ensemble inference engine, and a
large language model (LLM)-powered natural language explanation interface. DataMind was deployed and evaluated across three enterprise IDSS
applications: supply chain disruption prediction for a Dutch logistics consortium (EUR 2.1B annual revenue, 847 suppliers), dynamic pricing
optimisation for a Spanish e-commerce platform (4.7M daily transactions), and credit risk assessment for a French regional banking group (1.2M loan
applications, 2019–2024). DataMind achieved supply chain prediction F1-score of 0.923, pricing revenue improvement of 7.8%, and credit risk
AUC-ROC of 0.961 — surpassing production baselines by 18.4%, 3.2%, and 11.7% respectively. End-to-end decision latency averaged 340 ms from
data ingestion to recommendation delivery. These results establish DataMind as an accurate, low-latency, and explainable big data analytics
framework for enterprise intelligent decision support.
Keywords: ; ; ; ; ; ; ;