A new nonlinear causality test based on single multiplicative neuron model artificial neural network: a case study for Turkey’s macroeconomic indicators

Egrioglu E., Bas E., Cansu T., Kara M. A.

Granular Computing, vol.8, no.2, pp.391-396, 2023 (ESCI) identifier identifier

  • Publication Type: Article / Article
  • Volume: 8 Issue: 2
  • Publication Date: 2023
  • Doi Number: 10.1007/s41066-022-00336-z
  • Journal Name: Granular Computing
  • Journal Indexes: Emerging Sources Citation Index (ESCI), Scopus
  • Page Numbers: pp.391-396
  • Keywords: Causality tests, Artificial neural networks, Multiplicative neuron model, Particle swarm optimization, GRANGER CAUSALITY
  • Karadeniz Technical University Affiliated: No

Abstract

© 2022, The Author(s), under exclusive licence to Springer Nature Switzerland AG.Determining causal relationships is an important task in some scientific disciplines. The linear Granger causality tests have been commonly used in the literature. Although nonlinear causality tests have been proposed in the literature, there is no causality test based on a single multiplicative neuron model artificial neural networks. The contribution of this paper is proposing a new nonlinear causality test. In this study, a nonlinear causality test is proposed based on a single multiplicative neuron model artificial neural network which is trained by particle swarm optimization. The illustrations of the proposed tests are given using some of Turkey’s macroeconomic indicator time series. The test results show similar findings to the nonlinear causality test based on a multilayer perceptron but the proposed method produces smaller p values than the multilayer perceptron-based method.