This dissertation thesis deals with the prediction and optimization of real data using artificial neural networks methods in decision-making. Usage of artificial neural networks includes finding of various suitable types, topologies and learning algorithms of artificial neural networks to solve prediction of real-world data sets. Optimization solutions in the context of work focused on neural networks used in topology optimization learning algorithms and neural networks own calculation. The first part is a summary of current trends in applications development and the utilization of artificial neural networks. The review of the literature is the selection of the most common models of artificial neural networks for prediction of real data needs. Subsequently, the implementation of the defined application of artificial neural networks in support of decision-making process is made. On the basis of determining the value of the decision making process are defined methods for prediction of real data using classical methods. As classical methods are considered methods of statistical prediction models for time series. After that, the methods of artificial neural networks and their application on prediction task are described. In the next part, experiments involving the choice of type, topology, learning algorithms and optimization possibilities of artificial neural networks is tested on real-world data sets. The implementation presents the prediction of selected real-world economic values; namely a set of household expenditures and goods transport is used. Results obtained with artificial neural networks are compared with real value of the index and also with selected models of statistical time series prediction results.
| Identifer | oai:union.ndltd.org:nusl.cz/oai:invenio.nusl.cz:249237 |
| Date | January 2010 |
| Creators | Štencl, Michael |
| Source Sets | Czech ETDs |
| Language | Czech |
| Detected Language | English |
| Type | info:eu-repo/semantics/doctoralThesis |
| Rights | info:eu-repo/semantics/restrictedAccess |
Page generated in 0.0321 seconds