Knowledge Discovery

In the context of knowledge discovery, we believe these concepts can be linked as follow: Data are the lowest level of abstraction; researchers often speak about raw data to emphasize this fact. From data, models and patterns can be extracted, either automatically using data mining techniques or by humans using their conceptual, perceptual or visual skills respectively. The use of human intuition to come up with observations about the data is generally called insight, i.e., the act or outcome of grasping the inward or hidden nature of things or of perceiving in an intuitive manner. Patterns and models are not necessarily linked, even though some authors consider them as synonyms. One way to distinguish these two concepts is the following: patterns are directly attached to data or a sub-set of data; whereas models are more conceptual and are extra information that cannot necessarily be observed visually in the data. Further, the observation of some patterns can result in a model and inversely, the simulation of a model can result in a pattern. Hypotheses are derived from models and patterns. A validated hypothesis becomes information that can be communicated. Finally, information reaches the solid state of knowledge when it is crystallized, i.e., it reaches the most compact description possible for a set of data relative to some task without removing information critical to its execution.

[Bertini and Lalanne, 2009]

References

• [Bertini and Lalanne, 2009] Bertini, E. and Lalanne, D. 2009. Surveying the complementary role of automatic data analysis and visualization in knowledge discovery. In Proceedings of the ACM SIGKDD Workshop on Visual Analytics and Knowledge Discovery: integrating Automated Analysis with interactive Exploration (Paris, France, July 28 - 28, 2009). VAKD '09. ACM, New York, NY, 12-20