Technical solutions are described for improving a data-mining process for identifying a desired population of a dataset. An example computer-implemented method includes, receiving the dataset that includes a plurality of data dimensions. The method also includes determining a set of predictors for a target variable, where a predictor is a data dimension that is relevant to the target variable. The method also includes identifying from the set of predictors, a false predictor, where the false predictor is a data dimension that results in an empty set of the desired population. The method also includes removing the false predictor from the set of predictors used for the data-mining process for identifying the desired population of the dataset.