In wire electrical discharge machining (μ-WEDM), the stochastic nature of electrical discharges leads to complex surface textures characterized by pronounced peaks, valleys, and asymmetric height distributions, making accurate surface characterization essential. In this study, a hybrid framework based on surface roughness profile analysis and machine learning is proposed to distinguish machined surfaces produced under two different machining conditions and to infer the machining direction. A total of 400 two-dimensional roughness profiles were extracted from two machined surfaces, in the vertical and horizontal directions. Following profile preprocessing and the application of a Gaussian filter with a cutoff wavelength of 80 μm, four roughness parameters were extracted as input features for the KNN model. The performance of KNN algorithm was implemented with k-fold cross-validation for the horizontal and vertical directions. The results showed that using individual features resulted in classification accuracies ranging from 80.51% to 99.49% in the horizontal direction and from 63.00% to 100% in the vertical direction. In contrast, the simultaneous combination of the four features R_a, R_q, R_z, and R_sk resulted in a classification accuracy of 100% in both directions. Furthermore, comparison of the results obtained in the two directions showed that the differences in roughness parameters were more pronounced in the vertical direction, which can be used to infer the dominant machining direction. These findings demonstrate that integrating complementary surface roughness features with the KNN algorithm provides a simple and effective approach for distinguishing machining conditions and analyzing the machining direction in μ-WEDM surfaces.