Integrating Machine Learning with Cloud Analytics to Enhance Real-Time Business Intelligence

Abstract

Although business intelligence (BI) solutions historically analyzed near-real-time information, organizations increasingly seek to exploit real-time data. Dramatic reductions in the cost of data storage, cloud-enabled analytics, and investments to deliver streaming-ready information have created the potential to change the latency profile of BI systems. The introduction of machine learning (ML) in a cloud context represents another important opportunity—cloud infrastructure provides a family of services with rapidly decreasing cost and increasing ease of use that are optimized for ML and ML-related workloads. The concurrent desire to optimize the data-to-decision loop and the complementary nature of cloud analytics infrastructure and ML facilitate turning insights from active data into business actions, should that be required. However, these changes are not without challenges and require addressing the following questions: What architectures support integration of ML-driven information with BI? What considerations govern the design and operation of these architectures? Which real-world scenarios have achieved measurable performance improvements, shortening time to insight and supporting real-time data-driven decisioning? A range of publicly available real-time implementations across multiple industries demonstrate that these questions can be addressed, either wholly or in part, and that shortening time to insight improves BI.