The Review for Visual Analytics Methodology

Big data usage evolves from previously looking into the capacity of big data's descriptive and diagnostic perspectives into currently feeding the demands for predictive big data analytics. The needs come about due to organizations that crave predictive analytics capabilities to reduce risk, make intelligent decisions, and generate different customer experiences. Similarly, visual analytics play an essential role in understanding and fitting the analytics prediction in their business decision. Hence, the combination of descriptive, diagnostics and predictive within Visual Analytics emerges as a balanced field to provide understandable predictive insight. Due to the organizational demand and multi-discipline area, the approach to developing visual analytics is still uncertain in the Big Data Project Lifecycle from methodological perspectives. While there are a few potential methodological approaches that could be used for visual analytics, they are scattered across numerous academic research and industrial practice. To date, there is no coherent review and analysis of the work that has been explored specifically for Visual Analytics methodology. This paper reports on a review of previous literature concerning how Visual Analytics has been executed in the big data life cycle to address the gap. The review is organized in this study from three perspectives: i) general ICT -related methodology (e.g. SDLC, Agile, DevOps), ii) Data Science-related methodology (e.g. CRISP-DM, SEMMA, KDD) and iii) Visual Analytics-related methodologies in which each method will be benchmarked based on the Visual Analytics major part of reality, computer and human, in terms of its width, depth, and flows. This study found insufficiencies, non-specific and vague conditions in handling the Visual Analytics when using current methodological approaches based on the review conducted. The paper also highlights the Visual Analytics-related methodological review, which can shed some light on the approaches and ways of implementing analytics in the big data lifecycle, which can be beneficial for future studies in proposing a more comprehensive methodology for Visual Analytics in the big data lifecycle.

I. Introduction

Big data refers to massive, complex structured and unstructured data sets that are rapidly generated and transmitted from a wide variety of sources. It extracts value from the data and analyzes insights that lead to better decisions and strategic business moves. Big data usage evolves from descriptive, diagnostics, and more recently to predictions capabilities. As a result, Predictive Visual Analytics is currently in high demand for business and organization [1]. This is because organizations require predictive capabilities to reduce risk, make intelligent decisions, and generate different customer experiences. It attracts many industrial players to implement predictive analytics in their business [2]. In parallel, visual analytics play an essential role in understanding and fitting the analytics prediction in their business decisions. Hence, there is a need to embed prediction in visual analytics and becomes balanced to provide understandable predictive insights. When carefully executed, it can provide practical insights and predictions by analyzing current and historical data.