The comprehensive evaluation of students depends on several factors such as the intellectual quality,the ideological and moral quality,the physical and mental quality and the scientific and humanistic quality.The traditional evaluation of students is difficult to consider various aspects of students.To totally take account of various aspects of students,an improved selforganizing feature maps(SOFM) networks is proposed,which can map high-dimensional data into simple geometric relationships on a low-dimensional display effectively.The quantitative qualities of students are used as inputs to a SOFM.After giving some training,according to the location of the output neurons in the output layer,the students are finally classified into five categories by SOFM:excellent,good,general,less poor and poor.