Feature Selection and Error Tolerance for the Logical Analysis of Data

We present a new process of analyzing data to determine critical at-tributes in a classification problem. Our method allows for maximal error tolerance in the original data while producing a choice of simple Boolean formulas for classifying the data. We implement the method and test it on the Wisconsin Breast Cancer Database with classification accuracy over ninety percent. This process utilizes an existing dualiza-tion algorithm and the method of binarizing data with missing bits