The preliminary exploration on application of Machine Learning in the Middle-Infrared Spectrum analysis

Fourier mid-infrared spectrometers have a wide range of applications in chemical analysis, but it cannot be used as a module or efficient instrumentation in life and in a wider range of applications for the limitation of the database of infrared spectra and comparison methods. Therefore, we try to use the TensorFlow framework to build a model to make a preliminary exploration of trying to establish a general discriminant model. In the case of the binary classification task of oil and water, the model can achieve an average success rate of 0.92. Under the task of classifying 15 substances, the model can achieve an average success rate of 0.23.