Zographou, Athens: World Scientific and Engineering Academy and Society(WSEAS)
For many fuzzy data mining tasks, the information about the relationships between fuzzy attributes is a must. By a fuzzy simple correlation analysis, we can tell the strength of the linear relationship between two fuzzy attributes, and the direction of the relationship. And a fuzzy partial correlation analysis can provide us the relationship between two fuzzy attributes with other fuzzy attributes hold constant. These two fuzzy correlation analyses have been presented in previous works. Here, we turn to the discussion of the fuzzy semi-partial correlation, the correlation between two fuzzy attributes when the influences of other fuzzy attributes are removed from only one of the two attributes. A fuzzy semi-partial correlation analysis is greatly useful for us to choose the predictor fuzzy attributes in fuzzy prediction models. In this paper, the fuzzy semi-partial correlation is defined and derived with the membership grades of fuzzy attributes. Data from previous papers have been used to construct an analysis of fuzzy semi-partial correlation to a fuzzy prediction model.
WSEAS Transactions on Computers 5(12), pp.2970-2976