Output details

A hybrid heuristic approach for attribute-oriented mining

<15>We present a new extended heuristic clustering algorithm for mining inductive generalised rules. The novelty of our approach derives from the dynamic calculation of new thresholds from old ones using entropy and structural data features, and the computation of “interestingness” during generalisation. Comprehensive experimentation shows significant results: runtime is 2X as fast as existing methods, and interestingness is 4X higher than traditional algorithms. Algorithm has potential for parallelisation and can also be used as a pre-processing step to other mining algorithms that need input space reduction.