A global supply chain risk management framework: An application of text-mining to identify region-specific supply chain risks

Abstract

Nowadays global supply chains enable companies to enhance competitive advantages, increase manufacturing flexibility and reduce costs through a broader selection of suppliers. Despite these benefits, however, insufficient understanding of uncertain regional differences and changes often increases risks in supply chain operations and even leads to a complete disruption of a supply chain. This paper addresses this issue by proposing a text-mining based global supply chain risk management framework involving two phases. First, the extant literature about global supply chain risks was collected and analyzed using a text-based approaches, including term frequency, correlation, and bi-gram analysis. The results of these analyses revealed whether the term-related content is important in the studied literature, and correlated topic model clustering further assisted in defining potential supply chain risk factors. A risk categorization (hierarchy) containing a total of seven global supply chain risk types and underlying risk factors was developed based on the results. In the second phase, utilizing these risk factors, sentiment analysis was conducted on online news articles, selected according to the specific type of risk, to recognize the pattern of risk variation. The risk hierarchy and sentiment analysis results can improve the understanding of regional global supply chain risks and provide guidance in supplier selection.

Introduction

Global supply chains have been widely discussed to boost competitive advantages [1]. From an economic perspective, global supply chain management can create a win-win situation for associated stakeholders in that every role (or node) in the global supply chain network can obtain operational benefits when all stages related to the product/component production (e.g., design, manufacturing, assembly, testing, and marketing) is coordinated worldwide [2]. Although operating global supply chains has been considered as a common profit advantage in most industries, uncertain regional factors regarding global business and logistics such as government stability, insufficient infrastructure in a country, and trade barriers may make it difficult to have successful supply chain management [3], [4].

A global supply chain network can be interpreted as an integration of both internal and external stakeholders. This integration has positive impacts on corporate operations such as innovative strategy, product quality and profitability [5]. However, critical risks and challenges in global business operations may lead to the disruption of the global supply chain although suppliers in multiple tiers and various nations are fully coordinated to decrease the total supply chain cost. Christopher et al. [6] proposed a classification for global sourcing risks that include supply risk, environmental and sustainability risk, process and control risk, and demand risk. Among these risks, supply risk, process and control risk, and demand risk have been widely discussed in practice and academia because they are relatively easier to evaluate through quantitative methods common to domestic supply chain cases. From a global business dynamics vantage point, on the other hand, environmental and sustainability risks such as fluctuations in environmental protection legislations, taxation differences, and political stability are difficult to analyze and evaluate due to related uncertainties affected by local political, social and economic realities. This insufficient understanding of regional – often across national boundaries – differences and thus changes may significantly influence and disrupt the entire supply chain performance [6]. One significant current challenge in supply chain risk management is still about modeling and quantifying the risks [7]. A comprehensive risk categorization is needed to better understand the scope of global supply chain risk factors due to regional differences.

The above mentioned issues lead to the research questions: (1) What are the significant global supply chain risks due to regional differences? (2) How can such risks be identified and organized objectively? (3) How can the risk variation (factors and levels) patterns be recognized? Although there are previous studies on global supply chain risk management, most research applied case studies, operational-modelling or probability-related methodologies to categorize risks and address supply chain design problems [8], [9], [10]. Moreover, studies using qualitative literature review methods to classify global supply chain risks have challenges in validating due to their limited sample [11]. Therefore, a data-driven method is needed for researchers and practitioners to identify global supply chain risks and understand the significant risk patterns [12].

Artificial intelligence has been increasingly applied in supply chain risk management area; nevertheless, very few of these studies incorporated text analytics to extract useful information on supply chain risks [13]. To identify regional risks of a global supply chain by a data-driven approach, input data collection is the first and most important step to undertake. However, there is a great deal of text information on geographical differences and supply chain risks in published articles, business magazines, and news articles. General public is also able to receive information about natural disasters, government policies, and company operations from news media. Therefore, this study attempts to discover how this information can be exploited to reveal signals on global supply chain risks. Nowadays, such text-based information could be easily accessed through open portals or databases for academic purposes. Research databases such as Scopus, Engineering Village, and Google Scholar have a wide variety of peer-reviewed research articles, which are organized for retrieval by search algorithms. Published news articles are also stored on publisher’s website or database. Platforms such as Google News or Yahoo News collect the latest news articles from different mass communication media. These exemplify the abundance of information and sharing mechanisms, and demonstrates that online news media have already become a key element of social, economic, and cultural life worldwide [14]. This trend presents the possibility that such information can be sources of input to this research; peer-reviewed journal articles can help define important risk factors with their careful review processes, and online news articles aid in risk variation pattern recognition because of their content and time-specific relevance.

To analyze these articles directly for discovering valuable insights about regional supply chain risks, a text-oriented, unstructured data analysis method is required. Therefore, text mining is adopted to extract information from the input documents. Text mining is an approach that can analyze a large amount of documents (in a form of pre-processed text format) and extract valuable information. Text mining techniques have been applied to numerous areas such as political document analysis, energy and service industries, information technology, and healthcare service sectors due to their outstanding capability in analyzing unstructured data [15]. It is extensively applied in knowledge discovery, retrieval, and management. Artificial intelligence is also implemented in more advanced text-mining tools to analyze textual data to discover patterns and insights in unstructured texts [10].

This study addresses the above stated research questions by developing a text-mining-based risk management framework for global supply chains that consider up-to-date text information. This framework could provide guidance for companies to develop a resilient global supply chain considering the influences of regional differences and changes.

The remainder of this paper is organized as follows. Section 2 discusses previous studies of global supply chain risk management and data analytics (including text mining) applications. Section 3 demonstrates the proposed research framework. Section 4 presents the implementation and results of defining potential risk factors and risk variation pattern recognition incorporating sentiment analysis. A discussion on relevance to previous research and limitations of this study is also presented in this section. Finally, the summary of the proposed framework, highlights of contributions, and future research directions are included in Section 6.