Many wild nocuous fungi are similar to the edible in morphology and biological characteristic, which easily leads to serious food safety incident because it is difficult for farmers to distinguish them just by experience. The progress of wild edible production makes a great contribution to rural economy of Yunnan province where the yield and export volume are highest in China. Rapid authentication of wild edible fungi variety is beneficial for wild edible industry towards healthy development. Meanwhile, the authentication also contributes to the analysis of the genetic relationship between edible mushroom and their breeding. Seven kinds of fungi were collected from Yunnan and other seven origins around Yunnan. Fingerprint of caps and stipe were obtained with Fourier transforms infrared (FTIR) spectrometer, respectively. Cap model, stipe model, low-level data fusion model and mid-level data fusion were established using prepressed spectra according to low- and mid-level fusion strategy combined with decision trees, discriminant analysis, logistic regression classifiers, support vector machines, nearest neighbor classifiers and ensemble classifiers that every model was computed 10 times. The optimal classification algorithm was selected based on the accuracy of training set. Hierarchical cluster analysis (HCA) was executed using the mid-level fusion dataset to judge genetic relationship between seven fungi. The results indicated: (1) The best algorithm of caps, stipe and low-level fusion is linear discrimination that accuracy is 92.8%, 96.4%, and 97.6%, respectively. Subspace discriminant is the most optimal in mid-level fusion that accuracy is 100%. (2) The average accuracy of all samples is 93.61%, 95.54%, 96.99% and 99.88% based on the best model of stipe, cap, low-level data fusion and mid-level data fusion. The performance of mid-level fusion is better than other three models, which indicated that the model could distinguish the highly -similar samples by reducing the influence caused by their origins. (3) The result of HCA based on mid-level fusion dataset displayed that the distance between Boletus magnificus and B. edulis was very close, which showed their chemical information were similar and genetic relationship was close. (4) The result of HCA based on mid-level fusion dataset displayed that the distance between Boletus magnificus and Leccinum duriusculum was very long, which showed their chemical information were different and genetic relationship was inferior. In a word, mid-level data fusion strategy combining FTIR spectra of different parts, subspace discriminant and HCA could effectively distinguish different kinds of edible fungi and judge the genetic relationship, which is a novel method used for variety authentication and genetic relationship judgment of wild edible fungi.