Previous research has indicated there are some defects in the traditional immune-based algorithm of detecting abnormity. Two common defects of this are as follows: the detectors created by the algorithm cannot cover all Nonself-space, and the number of the detectors in the algorithm is too big to keep the detectors from covering each other. These defects are the main reasons why the algorithm is not efficient. In this paper, an algorithm, which can deal with mixed-type data and produce detectors in different sizes to improve the capacity of covering Nonself-space is proposed. The simulated and compared experiments show that this algorithm can preferably process mixed-type data and effectively improve the efficiency of generating detectors .