We applied neural networks to identification of chemical species. The input data corresponding to the spectra of X-ray photoelectron spectroscopy (XPS) and Auger electron spectroscopy (AES) were prepared using Gaussian patterns with noise component. The neural networks with a Kohonen's self-organized feature map and a back-propagation algorithm were used in this work. From the results, we found that the input patterns were able to be classified using the neural networks without any pre-treatments such as smoothing and so on. Therefore, neural networks are thought to be useful in chemical analysis.