There are abundant mixed data sets with various types of attributes in application fields. However, most multivariate data visualizations are only effective with simplex one data type. As for mixed data sets, the visualizations of them are usually dissatisfied. We present a data transformation technique for mixed data sets involving both numerical and categorical attributes. Firstly, every numerical attribute was categorized by clustering; then, all categorical attribute was quantified by Correspondence Analysis; finally, the transformed mixed data were presented in numerical data visualizations like Star Coordinates. Furthermore, aiming at those mixed data sets that have many attributes or the cardinality which is high, a set of cardinality reduction strategies were proposed to diminish the attributes number involved in computation to improve computational efficiency. Empirical studies show that the visualization of mixed data sets is easily-understandable and propitious for the user to discover the connotative information within; and that cardinality reduction strategies are highly memory-saving and time-efficient.