| Title |
Finansinių ir fizinių procesų prognozavimo modelių tyrimas ir taikymas / |
| Translation of Title |
Analysis and application of financial and physical forecasting models. |
| Authors |
Bernotas, Edvinas |
| Full Text |
|
| Pages |
54 |
| Keywords [eng] |
finance ; forecasting ; neural network |
| Abstract [eng] |
Nowadays investment became an important part of common people life. In order to invest successfully people need to anticipate market changes. Forecasting is one of the ways to make the decision. The main objective of this work is to develop neural network using different systems and compare it with other prediction models. Also to investigate how models predict different kind of data: real historical data of various sectors and virtual generated data. The analysis part of the work goes through artificial neural networks: their creation, training and validating. The neural networks were implemented and simulated using MatLAB and JOONE program packages. Root mean square error (RMSE) was selected to test the efficiency of the models. Results of experiments were compared with AR-ABS and RW methods. Results showed that neural network efficiency was worse than AR-ABS and RW methods. Although neural networks overtop other models in some cases and that leads to a conclusion that the proper use of neural networks could result to quite precise forecasts. |
| Type |
Master thesis |
| Language |
Lithuanian |
| Publication date |
2009 |
