Modeling European industrial production with multivariate singular spectrum analysis: A cross‐industry analysis

• Published date: 01 April 2018
• Date: 01 April 2018
• DOI: http://doi.org/10.1002/for.2508

Received: 2 February 2017 Revised: 30 August 2017 Accepted: 2 December 2017

DOI: 10.1002/for.2508

RESEARCH ARTICLE

Modeling European industrial production with multivariate

singular spectrum analysis: A cross-industry analysis

Emmanuel Sirimal Silva1Hossein Hassani2Saeed Heravi3

1Fashion Business School, London College

of Fashion, University of the Arts London,

London, UK

2Research Institute of Energy

Management and Planning, University of

Tehran,Tehran, Iran

3Cardiff Business School, Cardiff

University, Cardiff,UK

Correspondence

Emmanuel Sirimal Silva, FashionBusiness

School, London College of Fashion,

University of the Arts London, 272 High

Holborn, London WC1V 7EY,UK.

Email: e.silva@fashion.arts.ac.uk

Abstract

In this paper, an optimized multivariate singular spectrum analysis (MSSA)

approach is proposed to find leading indicators of cross-industry relations

between 24 monthly, seasonally unadjusted industrial production (IP) series

for German, French, and UK economies. Both recurrent and vector forecast-

ing algorithms of horizontal MSSA (HMSSA) are considered. The results from

the proposed multivariate approach are compared with those obtained via the

optimized univariate singular spectrum analysis (SSA) forecasting algorithm to

determine the statistical significance of each outcome. The data are rigorously

tested for normality, seasonal unit root hypothesis, and structural breaks. The

results are presented such that users can not only identify the most appropri-

ate model based on the aim of the analysis, but also easily identify the leading

indicators for each IP variable in each country. Our findings show that, for all

three countries, forecasts from the proposed MSSA algorithm outperform the

optimized SSA algorithm in over 70% of cases. Accordingly, this new approach

succeeds in identifying leading indicators and is a viable option for selecting the

SSA choices Land r, which minimizes a loss function.

KEYWORDS

forecasting, industrial production, leading indicators, multivariate SSA, singular spectrum analysis

1INTRODUCTION

The introduction of the nonparametric time series analy-

sis and forecasting technique of singular spectrum analysis

(SSA) is closely associated with the work of Broomhead

and King (1986a, 1986b). Since then, SSA has progressed

rapidly and transformed itself into a powerful technique

that is increasingly exploited for providing solutions to

real-world problems in a variety of different fields; see,

for example, Gong, Song, He, Gong, and Ren (2017),

Merte (2017), Yu, Li, and Zhang (2017), Mahmoudvand

and Rodrigues (2017), Khan and Poskitt (2017), Hassani,

Ghodsi, Silva, and Heravi (2016), Hassani, Silva, Anton-

akakis, Filis, and Gupta (2017), Hassani, Webster, Silva,

and Heravi (2015), Lai and Guo (2017), Ghodsi, Silva, and

Hassani (2015), and Silva and Hassani (2015). Few reasons

underlying this augmented usage of SSA can be partly

attributed to its nonparametric nature. This implies that

the parametric assumptions relating to normality, station-

arity, and linearity do not apply when modeling with SSA.

In addition, SSA is popular for its sound filtering capabil-

ities, which enable signal extraction from any given time

series and facilitate a richer analysis (see, e.g., Ghodsi et al.,

2015). Furthermore, over the years, SSA has been suc-

cessful in analyzing and forecasting both stationary and

nonstationary time series, which is of utmost importance

to the financial and economic fields where uncertainty and

recessions are of major concern.

As noted above, SSA (used to refer to the univariate

version) is widely exploited at present and there exists a

Journal of Forecasting. 2018;37:371–384. wileyonlinelibrary.com/journal/for Copyright © 2018 John Wiley & Sons, Ltd. 371