| Abstract [eng] |
The goal of this research work is to adapt the necessary data preparation methods and to create a classification model for Raman spectroscopy data of various types of meat, capable of reliably recognizing the classes of data. The tasks of the work are to get acquainted with the existing methods of Raman spectrum analysis, to investigate the mostly used machine learning methods and to select the most appropriate algorithm that could be used for future classification of this type of data. During the research work the Raman spectrum data type identification experiments were carried out using a support vector machine algorithm, artificial neural networks, and convolutional neural networks. High accuracy was achieved with all the methods studied. The best results were obtained using the artificial neural network method. By this method, the classification model has reached a precision of 96.85%. It was also found that the neural network based models works well without applying baseline correction. |
