Fluorescence correlation spectroscopy is an established technique in single-molecule spectroscopy used for observing temporal fluctuations of the fluorescence intensity of marked biomolecules. Fluctuations are temporally autocorrelated which enables further analysis and measurement of relevant system parameters. Examples of observables are molecule concentrations, diffusion coefficients, etc. In this master’s thesis an unbound Alexa647 dye molecule was used as source of fluorescence. By using an approximate theoretical function, which incorporates diffusion dynamics and triplet dynamics of the dye, dimensions of the observational volume of the confocal microscope were calculated. The latter parameters are then fixed in the theoretical model which enabled determination of diffusion time and diffusion constant of the dye. As an alternative method, machine learning was used to process autocorrelation curves. Specifically, a neural network was trained on a dataset that included data of known parameters. After the model was trained, it was used to predict diffusion coefficients of the data in the test set which was not used during the model training. System parameters were measured directly from the data. Doing it that way enables us to potentially avoid systematic errors which come from optical artifacts of the experimental apparatus because approximate theoretical functions do not take optical artifacts into account. The results of the model are compared with the results obtained in the theoretical function fitting procedure. Obtained results signalled that even though the training set used for model training was small, the machine learning model was at least as successful in predicting diffusion coefficients as was the theoretical function fitting procedure.