The invention discloses a mixing attribute data flow cluster method based on reinforcement cluster
edge detection of a grid, comprising steps of 1) a grid pre-
processing process, 2) an online grid maintenance process, and 3) an off-line cluster process. The grid pre-
processing process includes steps of dividing a d dimension
data space where the data object is positioned, dividing each dimension of
numeric data into P equally-divided sections according to the size of the grid
granularity due to the fact that the mixing attribute comprises a
numeric value attribute and the classification attribute, dividing classification data of each dimension according to the possible number of the value in the domain and dividing a
data space into a plurality of measure polytopes which are not mutually intersected, wherein each rectangle grid unit expression is S1,J1*S2,j2*...*Sd,jd, wherein the attributes Si,(i<d) is an attribute on the
data space S, and the subscript ji expresses the section obtained on the dimension of the Si. The invention provides a mixing attribute data flow cluster method based on reinforcement cluster
edge detection of a grid which is high in cluster quality and strong in
processing rim network.