Quantisation of Continous Valued Data through Scatter Search for Machine Learning Applications

摘要

This paper discusses a two step procedure for quantisation of continuous valued data: i) finding optimal threshold value that separates two consecutive quantisation levels ii) identifying optimal location of partitions. Application of scatter search for the second step of the procedure is explored. A formulation of the problem statement is first presented. A detailed discussion of various procedures of scatter search in the context of this problem is then presented.