A requirement engineering model for big data software

Most prevailing software engineering methodologies assume that software systems are developed from scratch to capture business data and subsequently generate reports. Nowadays, massive data may exist even before software systems are developed. These data may also be freely available on Internet or may present in silos in organizations. The advancement in artificial intelligence and computing power has also prompted the need for big data analytics to unleash more business values to support evidence-based decisions. Some business values are less evident than others, especially when data are analyzed in silos. These values could be potentially unleashed and augmented from the insights discovered by data scientists through data mining process. Data mining may involve overlaying and merging data from different sources to extract data patterns. Ideally, these values should be eventually incorporated into the information systems to be. To realize this, we propose that software engineers ought to elicit software requirements together with data scientists. However, in the traditional software engineering process, such collaboration and business values are usually neglected. In this paper, we present a new requirement engineering model that allows software engineers and data scientists to discover these values hand in hand as part of software requirement process. We also demonstrate how the proposed requirement model captures and expresses business values that unleashed through big data analytics using an adapted use case diagram.

I. Introduction

The focus on data in information systems is rapidly increasing nowadays due to the vast amount of data generated by software systems. Youtube, for example, had more than a billion users in 2015, who watched 4 billion videos every day and uploaded 300 hours of video every minute [1]. In this hurricane of data generated from the usage of software applications, software engineers need to be aware of what Big Data is, how to develop Big Data software and what the requirements of Big Data software are. As a general definition, big data are data with 5V's; variety, velocity, volume, veracity, and value [2]. Taking big data into account in software engineering will enhance the quality of software and exploit the values of such data that would otherwise be wasted. When appropriately exploited, the business values (or simply values) of big data can affect the systems positively in several ways. For example, it enhances the quality of big data software and helps make accurate predictions regarding day to day business which may assist in informed decisions to maximize the quality of services and profit. Moreover, with the advent of Internet of things (IoT), cloud computing and machine learning techniques have also indirectly enabled software systems to process big data, and subsequently unleash and exploit business values by learning from their previous activities in order to manage future activities better, and subsequently optimize their business goals. This is evident in the case study explained in section IV.