A design proto‐pattern for continuously evaluated forecasting in IBM® InfoSphere® Streams. (26th February 2015)

• Record Type: Journal Article
• Authors: Oliver, Helen; McSharry, Patrick
• Is Part Of: Software, practice & experience. Volume 46:Number 5(2016)
• Page Start: 645
• Page End: 684
• Publication Date: 2015-02-26
• Subjects: design patterns -- forecasting -- stream computing -- IBM® InfoSphere® Streams -- forecast performance evaluation -- near real time; Computer software -- Periodicals; Computer programming -- Periodicals; Computer programs -- Periodicals; 005.3
• DOI: 10.1002/spe.2316
• Languages: English
• ISSNs: 0038-0644

Abstract

Summary: Design patterns relieve much of the difficulty of developing software solutions. As sources and volumes of generated data have increased in recent times, stream computing has emerged as a new paradigm for acquiring and harnessing the data's rich potential to provide insight in near real time. One prominent platform for processing streaming data is IBM® InfoSphere® Streams, but design patterns for the platform are relatively scarce. This article proposes a prototypical application design pattern for continuously evaluated forecasting on the IBM InfoSphere Streams platform. Towards demonstrating the prototypical pattern, the application was implemented in three real‐world scenarios: wind power, sentiment analysis, and stock market data. The purpose of this article is to specify the design proto‐pattern and describe its variants in each different implementation. We believe that the proto‐pattern will be applicable in IBM InfoSphere Streams to any data stream where forecasts over multiple forecast horizons are required. Copyright © 2015 John Wiley & Sons, Ltd.