One of the most insightful definitions of data mining states that to be truly successful data mining should be “the nontrivial process of identifying valid, novel, potentially useful, and ultimately comprehensible knowledge from databases” (Fayyad,1997). However, to achieve this aim we need to develop the so called "interestingness" measures to overcome certain issues.
The main issue is that the data mining process can generate many hundreds and often thousands of patterns from data. The task for the data miner then becomes one of determining the most useful patterns from those that are trivial or are already well known to the organization.
The problem, as Gaines points out this is not the easiest metric or criterion to implement in a data mining system because the most interesting discoveries are those that are unforeseen and surprising i.e. we all know a novel pattern when we see one but find it difficult to provide the guidance necessary (i.e. to be able to articulate this to a human never mind a computer) to discover one (Gaines,1996).
Gaines, B. (1996), Transforming rules and trees, in U. Fayyad, G. Piatetsky-Shapiro, P. Smyth and R. Uthursamy, eds, ‘Advances in Knowledge Discovery and Data Mining’, AAAI-Press, pp. 205–226.
Fayyad, U. and Stolorz, P. (1997), ‘Data mining and kdd: Promise and challenges’, Future Generation Computer Systems 13 (2-3), 99–115
McGarry, K. A survey of Interestingness Measures for Knowledge Discovery, Knowledge Engineering Review Journal, Vol 20, No 1, 39-61. 2005.