Physical Time Series Prediction Using Dynamic Neural Network Inspired by the Immune Algorithm

• Published in: Adaptive and Intelligent Systems pp. 152 - 161
• Main Authors: Hussain, Abir Jaafar, Al-Askar, Haya, Al-Jumeily, Dhiya
• Format: Book Chapter
• Language: English
• Published: Cham Springer International Publishing 2014
• Series: Lecture Notes in Computer Science
• Subjects: and physical time series prediction; Recurrent neural network; self organised neural network
• ISBN: 9783319112978; 331911297X
• ISSN: 0302-9743; 1611-3349
• DOI: 10.1007/978-3-319-11298-5_16

Time series analysis is a fundamental subject that has been addressed widely in different fields. It has been exploited and used in different scientific fields for example, natural, biomedical, economic and industrial data as well as financial time series. In this paper, we consider the application of a novel neural network architecture inspired by the immune algorithm and the recurrent links for the prediction of Lorenz and earthquake time series by exploiting the inherent temporal capabilities of the recurrent neural model. The performance of this network is benchmarked against “traditional”, rate-encoded, neural networks; a Multi-Layer Perceptron network, a Jordan and an Elman neural network as well as the self organized neural network inspired by the immune algorithm. The results indicate that the inherent temporal characteristics of the recurrent links network make it extremely well suited to the processing of time series based data.