Elsevier

Journal of Econometrics

Volume 81, Issue 1, November 1997, Pages 273-280
Journal of Econometrics

Recognizing changing seasonal patterns using artificial neural networks

https://doi.org/10.1016/S0304-4076(97)00047-XGet rights and content

Abstract

In this paper we propose a graphical method based on an artificial neural network model to investigate how and when seasonal patterns in macroeconomic time series change over time. Neural networks are useful since the hidden layer units may become activated only in certain seasons or periods, and since this activity can be stepwise or smooth. The graphical method is based on the partial contribution of the hidden layer units to the overall fit. We apply our method to quarterly Industrial Production in France and Netherlands.

References (11)

There are more references available in the full text version of this article.

Cited by (48)

  • Forecasting Bitcoin with technical analysis: A not-so-random forest?

    2023, International Journal of Forecasting
    Citation Excerpt :

    In technical analysis applications, Hans and van Griensven Kasper (1998), Gençay and Stengos (1998), and Fernández-Rodríguez et al. (2000) indicated that utilizing past buy–sell signals of foreign exchange rates and stock prices in a feedforward ANN improves profitability and out-of-sample generalizations. Franses and Draisma (1997) proposed a method based on ANNs to investigate when and how seasonal patterns in macroeconomic time series change over time. Yang et al. (1999) used probabilistic ANNs in bankruptcy prediction.

  • How macro-variables drive crude oil volatility? Perspective from the STL-based iterated combination method

    2022, Resources Policy
    Citation Excerpt :

    Thus, there are important economic mechanisms between them, which is the basis of forecasting COPV. Second and more important, significantly seasonal behavior has been observed in these macro variables (Sealey, 1977; Franses and Draisma, 1997; Tiwari et al., 2014). Hence, whether and how this important characteristic of the macro variables can be employed in modeling the forecasting of COPV?

  • Hybrid modeling in the predictive analytics of energy systems and prices

    2020, Applied Energy
    Citation Excerpt :

    The study by [23] used an ANN model to forecast electric load. The popularity of ANN models has been increasing because they have ability to model a time series data with seasonal patterns without any pre-processing[24–27]. Moreover, ANN models not only can deal with complex patterns, but they also perform well with missing and incomplete data [28].

  • Forecasting tourism demand with denoised neural networks

    2019, Annals of Tourism Research
    Citation Excerpt :

    However, it is noteworthy that there have been some authors who reported that NN can model the trend and seasonal effects in time series without the need for denseasonaling (Hamzacebi, 2008). See for example, Franses and Draisma (1997) and Alon, Qi, and Sadowski (2001). Tseng, Yu, and Tzeng (2002) sought to create a hybrid model which combined NN back propagation (BP) with a seasonal ARIMA (SARIMA) model and found this hybrid model could outperform two other NN models and the univariate SARIMA model when faced with seasonal time series.

  • Data driven model for sonic well log prediction

    2018, Journal of Petroleum Science and Engineering
    Citation Excerpt :

    Most importantly, there is no set rule on how an ANN derived model should be designed or developed to get the best results (Adedigba et al., 2017). For this reason, it is recommended to adopt simple ANN architecture except in cases where complex correlations will better describe the existing structure between the input and target values (Franses and Draisma, 1997). Note that increasing the number of input vectors or data is not synonymous with a better prediction as it is possible to over fit the data.

View all citing articles on Scopus

This paper was presented at an Institute of Economics seminar in Aarhus (November 1994) and at the (EC)2 conference in Berlin (December 1994).

1

We thank the participants for their helpful comments. Special thanks go Herman Bierens, Svend Hylleberg, Johan Kaashoek, Halbert White, and to Helmut Lütkepohl (the Associate Editor) and two anonymous referees for their helpful suggestions.

2

The first author thanks the Royal Netherlands Academy of Arts and Sciences for its financial support.

View full text