Continuous attribute reduction method based on an automatic clustering algorithm and decision entropy

The data in the process of power plant are mostly continuous. Aimed at the problems existing in the discretization of the continuous attributes with associated relationship, a strategy combining k-means clustering algorithm and the degree of compatibility of the decision table was proposed. The degree of compatibility was taken as judgment conditions to determine whether a new cluster type is added, while ensuring that the degree of compatibility can be accepted. This method take full account of the relationship among variables. The automatic clustering is realized when discretizing the continuous attributes, which makes the clustering results effective and reasonable. An attribute reduction algorithm based on approximate decision entropy is used to extract the main relations of the system, and the dimensionality of the data set can be effectively reduced. The improved scheme is applied to continuous data of main steam temperature system of thermal power plant, and the validity of the method is verified.