Services on Demand
Journal
Article
Indicators
- Cited by SciELO
- Access statistics
Related links
- Cited by Google
- Similars in SciELO
- Similars in Google
Share
Prospectiva
Print version ISSN 1692-8261
Abstract
MERCADO POLO, Darwin; PEDRAZA CABALLERO, Luis and MARTINEZ GOMEZ, Edinson. Comparison of Neural Network applied to prediction of Time Series. Prospect. [online]. 2015, vol.13, n.2, pp.88-95. ISSN 1692-8261. https://doi.org/10.15665/rp.v13i2.491.
The main objective of the present work is to compare artificial neural networks (ANN) Multilayer Perceptron (MLP) and radial basis function (RBF) applied to time series prediction. The learning algorithm used was the resilient backpropagation for MLP network and a combination of the K-Means algorithm and pseudoinverse matrix method for the RBF. The implementation of the RNA was performed using a client-server-based system architecture, anticipating a future integration with real-time applications. For the evaluation of RNA, datasets with different characteristics and amount of data was used. According to the results, it can be concluded that the use and integration of computational intelligence techniques in web systems, it is preferable to use the RBF, because you get better runtimes. It is also important to note that response of both types of neural networks obtained similar results.
Keywords : Artificial Neural Networks; Forecasting; Time Series; Multilayer Perceptron; Radial Basis Functions.