BKAOS: An Extension of KAOS for Big Data Projects

The success of software projects is greatly impacted by requirements that are corrected on time. In the development process, fixing requirements errors early in the requirements engineering (RE) phase saves a significant amount of effort. Requirements, however, must be carefully manufactured; they are not just found. This holds true for systems in general and is particularly crucial for big data projects due to their specific characteristics, such as volume, variety, volatility, and the execution time, etc. Unfortunately, there is very little research on eliciting requirements specifically tailored to big data. In this study, we analyze existing literature to identify big data characteristics that are not adequately supported by traditional RE methods. We propose the BKAOS (big data KAOS) method, which is an improved form of the famous KAOS, to close this gap. Through 15 illustrative scenarios, we show the usefulness of BKAOS. We further verify the integrity of the method by generating a Bigraphs-based description for both BKAOS and KAOS. Our results show that BKAOS is more appropriate for eliciting requirements in big data projects than KAOS, which results in more precise requirement elicitation, lower effort levels overall, and easier data analysis.