The classified feasibility of Malus asiatica fruit by using near-infrared spectroscopy (VIS/NIR spectrometer) and softening index during different shelf life was analyzed. A mathematical model was established by combining partial least square (PLS) and standard normalized variate (SNV) which was regarded as the best pre-processing technology. The determination coefficients (R2) of calibration set and prediction set were 0.847 and 0.813 respectively, which illustrated that there was a high correlation between spectrum and the softening index. It showed that the softening index could be used to classify Malus asiatica samples during different shelf life. The LS-SVM and PLS-LDA were applied to build classification models, respectively. The results indicated that non-linear LS-SVM model was more suitable for classification of Malus asiatica samples than linear PLS-LDA model. The average correct recognition rate and average correct rejection rate were above 94%.