Developing a Generic Framework for Constructing Visual Data Mining Tools

A recent study has estimated that the "digital universe " has grown by a factor of nine since 2005 to 1.8 zettabytes (1000 million terrabytes) of data in 2011. Despite this improving ability to store data, it has been suggested that there has been little corresponding improvement in the availability of knowledge. This predicament arises since our capacity to generate and store data by far exceeds our current ability to explore and analyse these vast volumes. Visual approaches to data mining have attempted to address this problem by exploiting human perception in the discovery of interesting and meaningful patterns in abstract data. Through visualisation data may be explored directly and intuitively to gain new insights and to draw cogent conclusions. Visualisation also helps to promote confidence and trust in data mining models and to justify actions taken in light of those models. Although visualisation has proved an effective technique in this regard, as yet little other than guidelines exist to support the development of visual data mining tools. In view of this shortcoming, this paper explores the rationale for constructing a generic framework for the development of visual data mining tools that would attempt to tightly integrate visualisation with data mining techniques. This paper also identifies a programme of research that will lead to the construction of such a framework