The diamondback moth (DBM), Plutella xylostella (L.) is considered as the most destructive pest of Brassicaceae crops world-wide. Its migratory capacities and development of insecticide resistance in many populations leads to more difficulties for population management. To control movement of populations and apparitions of resistance carried by resistant migrant individuals, populations must be identified using genetic markers. Here, seven different ISSR markers have been tested as a tool for population discrimination and genetic variations among 19 DBM populations from Canada, USA, Brazil, Martinique Island, France, Romania, Austria, Uzbekistan, Egypt, Benin, South Africa, Réunion Island, Hong Kong, Laos, Japan and four localities in Australia were assessed. Two classification methods were tested and compared: a common method of genetic distance analyses and a novel method based on an advanced statistical method of the Artificial Neural Networks' family, the Self-Organizing Map (SOM). The 188 loci selected revealed a very high variability between populations with a total polymorphism of 100% and a global coefficient of gene differentiation estimated by the Nei's index (Gst) of 0.238. Nevertheless, the largest part of variability was expressed among individuals within populations (AMOVA: 73.71% and mean polymorphism of 94% within populations). Genetic differentiation among the DBM populations did not reflect geographical distances between them. The two classification methods have given excellent results with less than 1.3% of misclassified individuals. The origin of the high genetic differentiation and efficiency of the two classification methods are discussed.