Incursions of the Queensland fruit fly Bactrocera tryoni (Qfly) into areas without permanent Qfly populations present serious threats to the Australian and New Zealand horticultural industries. Identifying the origins of recent incursions will help reduce future threats by enabling the targeting of problematic incursion routes for more stringent quarantine protocols. Here we present an analytical framework based on supervised and unsupervised machine learning to identify the origins and recent population history of incursion individuals. Our framework is based on a recently developed reference dataset of genome-wide markers for 35 Qfly populations from across the ranges of Qfly and the related taxon Bactrocera aquilonis (NTfly). We apply our framework to recent incursions into New Zealand, Tasmania and South Australia. Two distinct Qfly sources were identified for incursions into New Zealand (total 18 individuals), one from the east coast of Australia and one from New Caledonia. All eight recent incursion collections analysed (total 85 individuals) from South Australia and Tasmania most likely originated from just one of six clusters of populations in our reference database, Qfly from the east coast of Australia. None were found to originate from clusters containing NTfly or Qfly/NTfly hybrids in the Northern Territory or north Western Australia. Several, but not all, of the collections showed signals of small founding population size and two Tasmanian collections each included individuals apparently derived from three different sources within the east coast of Australia. In total, several more incursion events were detected than previously known, although some were founded by relatively few individuals.