A neural network of the multiple-layer perceptron (MLP) type, named CERNEUR, has been created for the task of analysing the charge exchange recombination data from DIII-D for the purpose of providing control-room ion temperatures and rotation velocities between shots and, in the future, to provide initial guesses for the standard curve-fitting code. The perceptron is trained using back propagation which has been modified by the 'General Adaptive Recipe' (GAR). The input to CERNEUR is the output from the 250 pixel Reticon arrays which are the detectors in the CER (charge exchange recombination) spectrometers. The data are normalized and frame-shifted. The normalization and frame-shift factors are also included in the CERNEUR input data. The target training output data are obtained from the results from CERFIT, which is a code which fits Gaussians to the various peaks in the spectrum and returns amplitudes, temperature and locations. CERNEUR provides very useful 'control-room' in-between shot analysis of the rotation velocity and ion temperature profiles.