Contextualization of Big Data Quality: A framework for comparison

TL;DR

This paper proposes a comprehensive framework for incorporating contextual information into big data quality assessment, enhancing the accuracy and usefulness of quality evaluation methods.

Contribution

It introduces a novel context classification framework with new features and dimensions for big data quality assessment, improving upon existing models.

Findings

The model is more understandable and comprehensive.

It is richer and more useful than existing models.

Initial evaluation shows improved performance.

Abstract

With the advent of big data applications and the increasing amount of data being produced in these applications, the importance of efficient methods for big data analysis has become highly evident. However, the success of any such method will be hindered should the data lacks the required quality. Big data quality assessment is therefore a major requirement for any organization or business that use big data analytics for its decision making. On the other hand, using contextual information is advantageous in many analysis tasks in various domains, e.g. user behavior analysis in the social networks. However, the big data quality assessment has benefited less from this potential. There is a vast variety of data sources in the big data domain that can be utilized to improve the quality evaluation of big data. Including contextual information provided by these sources into the big data quality assessment process is an emerging trend towards more advanced techniques aimed at enhancing the performance and accuracy of quality assessment. This paper presents a context classification framework for big data quality, categorizing the context features into four primary dimensions: 1) context category, 2) data source type that contextual features come from, 3) discovery and extraction method of context, and 4) the quality factors affected by the contextual data. The proposed model introduces new context features and dimensions that need to be taken into consideration in quality assessment of big data. The initial evaluation demonstrates that the model is more understandable, more comprehensive, richer, and more useful compared to existing models.