To investigate differences in flavor quality among dried banana chips processed using hot air drying (HAD), vacuum freeze drying (VFD), and heat pump drying (HPD), and to establish predictive discrimination models for drying methods based on flavor quality indicators, this study analyzed flavor characteristics using an electronic nose (E-nose) and gas chromatographymass spectrometry (GC-MS). Relative odor activity values (ROAVs) were used to identify key aroma-active compounds. Principal component analysis (PCA) and linear discriminant analysis (LDA) were applied for dimensionality reduction and feature extraction, and a backpropagation neural network (BPNN) model was constructed for drying method discrimination and prediction. The results demonstrated that E-nose analysis revealed a distinctive response of HAD samples at the W3C sensor compared with the other drying methods. Seventeen key aroma compounds, including 11 esters, were identified based on ROAV analysis. Cluster heatmap analysis indicated higher flavor similarity between the VFD and HPD samples. Among the established models, the LDABPNN model achieved the highest discrimination performance (R2=0.844 4), outperforming the PCA-BPNN (R2=0.766 9) and the standalone BPNN model (R2=0.703 6), and completely distinguishing HAD samples from the other groups. The flavor profiles of dried banana chips produced using different drying methods were systematically characterized, and an LDA-BPNN model based on flavor quality indicators demonstrated strong predictive performance for drying method discrimination. These findings provide a technical reference for optimizing drying processes and improving the quality of dried banana chip products.