With the KNM gold ore deposit as an example, the authors carried out qualitative discrimination of nuclear geophysical anomaly by using discriminant analysis composed of canonical discriminant and stepwise discriminant analysis. Discriminant analysis was employed for gold prospecting, while SPSS software was used for data analysis. It is found that, with these two means, accurate characterization of barren anomalies reaches 83.3% and 66.7% respectively, while accurate characterization of ore-related anomalies reaches 89.7% and 91.4% respectively.