In order to discriminate adulterated goat milk quickly and objectively, a set of portable electronic tongue detection system was exploited, and a new method of fast identification is developed. When detected in the system, the sample solution was first scanned to obtain the "fingerprint" information of adulterated goat milk, and then the discrete wavelet transform (DWT) was used to obtain the characteristics of the "fingerprint" data. On this basis, the principal component analysis (PCA) was used to determine the quality of goat milk with different adulteration ratio. Particle swarm optimization extreme learning machine (PSO-ELM) was applied to quantitatively predict goat milk with different adulteration proportions. According to the experimental data, PCA could distinguish six kinds of goat milk with different adulteration ratios up to 100%, and it had a good effect on distinguishing adulterated goat milk. In order to realize the quantitative prediction of goat milk with different adulteration ratios, the fitting curve of PSO-ELM goat milk purity prediction model was very close to the measured curve, so the PSO-ELM method was used to establish the quantitative prediction model of goat milk purity with high prediction accuracy. This study might provide new ideas and technical support for qualitative identification and quantitative prediction of adulterated goat milk.