IRIS: A Goal-Oriented Big Data Business Analytics Framework

Abstract

Big data analytics is the hottest new practice in Business Analytics today. However, recent industrial surveys find that big data analytics may fail to meet business expectations because of lack of business context and lack of expertise to connect the dots, inaccurate scope and batch-oriented Hadoop system. In this dissertation, we present IRIS – a goal-oriented big data analytics framework for better business decisions, which consists of a conceptual model that connects a business side and a big data side, providing context information around the data, an evidence-based evaluation method which enables to focus the most effective solutions, a process on how to use IRIS framework and an assistant tool using Spark, which is a real-time big data analytics platform. In this framework, problems against business goals of the current process and solutions for the future process are explicitly hypothesized in the conceptual model and validated on real big data using big analytics queries. As an empirical study, a shipment decision process is used to show how IRIS can support better business decisions in terms of comprehensive understanding both on business and data analytics, high priority and fast decisions. Additionally, at the core of Big Data lies data, which is essential for supporting business analytics in gaining insights about business practices towards making better business decisions. The quality of business analytics inevitably depends on the kinds of individual data and relationships between the data, which should all be defined in a data model. A poor data model can lead to omissions or commissions of important business considerations, likely resulting in bad business decisions. However, there is little work on systematically and rationally developing a big data model for better supporting business analytics, especially in the presence of a variety of sources and types of data that are increasingly becoming available and useful. In this dissertation, we propose three notions of big data model quality – relevance, comprehensiveness and relative priorities with a goal-oriented approach to building such qualities in a big data model. In this goal-oriented approach, alternatives in big data models are explored and selected for validating potential problems and solutions, while also achieving business goals. An empirical study has been conducted on the shipping decision process of a world-wide retail chain, to gain an initial understanding of the applicability of this approach. Finally, many software systems are being developed to help with business processes, which typically involve a number of (human) tasks in achieving organizational goals. However, aligning a software system well with its intended business process has been challenging, since the tasks in a business process usually lack formal definitions and can be performed via multiple different allocations of resources. In this dissertation, we propose a goal-oriented transformational approach to deriving use cases, as requirements on the software system, from a business process which is modeled in BPMN (Business Process Model and Notation). In this approach, a business process is modeled not only in terms of the functionally-oriented BPMN but also non-functional business goals, and the target software requirements are also modeled in terms of functionally-oriented use cases together with non-functional requirements. Those tasks to be performed by a software system are transformed into use cases, in consideration of multiple alternative interpretations of business tasks, different allocations of software functionality and the granularity of the target requirements guided via similarity and granularity. Additionally, an intermediate model is utilized in the 2-step transformation process to deal with the ontological gap and the many-to-many relationships between the source and the target. This process is facilitated by context-aware transformation rules and a supporting tool. A study of a quote flow business process shows that our goal-oriented transformational approach helps produce more cohesive, correct and comprehensive use cases.