Insufficient data on the origins of the first introduced propagule and the initial stages of invasion complicate the reconstruction of a species’ invasion history. Phylogeography of the native area profoundly shapes the genomic patterns of the propagules on which subsequent demographic processes of the invasion are based. Thus, a better understanding of this aspect helps to disentangle native and invasive histories. Here, we used genomic data together with clustering methods, explicit admixture tests combined with ABC models and Machine Learning algorithms, to compare patterns of genetic structure and gene flow of native and introduced populations, and infer the most likely invasion pathways of the highly invasive freshwater fish Pseudorasbora parva. This species is the vector of a novel lethal fungal-like pathogen (Sphaerothecum destruens) that is responsible for the decline of several fish species in Europe. We found that the current genetic structuring in the native range of P. parva has been shaped by waves of gene flow from populations in southern and northern China. Furthermore, our results strongly suggest that the genetic diversity of invasive populations results from recurrent global invasion pathways of admixed native populations. Our study also illustrates how the combination of admixture tests, ABC and Machine Learning can be used to detect high-resolution demographic signatures and reconstruct an integrative biological invasion history.