A practical, collaborative approach for modeling big data analytics application requirements

Abstract

Data analytics application development introduces manychallenges including: new roles not in traditional softwareengineering practices e.g. data scientists and data engineers; useof sophisticated machine learning (ML) model-based approaches;uncertainty inherent in the models; interfacing with models tofulfill software functionalities; deploying models at scale andrapid evolution of business goals and data sources. We describeour Big Data Analytics Modeling Languages (BiDaML) toolset tobring all stakeholders around one tool to specify, model anddocument big data applications. We report on our experienceapplying BiDaML to three real-world large-scale applications.Our approach successfully supports complex data analyticsapplication development in industrial settings.