Rigorous Measurement Model for Validity of Big Data: MEGA Approach

Abstract

Big Data is becoming a substantial part of the decision-making processes in both industry and academia, especially in areas where Big Data may have a profound impact on businesses and society. However, as more data is being processed, data quality is becoming a genuine issue that negatively affects credibility of the systems we build because of the lack of visibility and transparency of the underlying data. Therefore, Big Data quality measurement is becoming increasingly necessary in assessing whether data can serve its purpose in a particular context (such as Big Data analytics, for example). This research addresses Big Data quality measurement modelling and automation by proposing a novel quality measurement framework for Big Data (MEGA) that objectively assesses the underlying quality characteristics of Big Data (also known as the V's of Big Data) at each step of the Big Data Pipelines. Five of the Big Data V's (Volume, Variety, Velocity, Veracity and Validity) are currently automated by the MEGA framework. In this paper, a new theoretically valid quality measurement model is proposed for an essential quality characteristic of Big Data, called Validity. The proposed measurement information model for Validity of Big Data is a hierarchy of 4 derived measures / indicators and 5 based measures. Validity measurement is illustrated on a running example.