Machine learning in international business

Abstract

In the real world of international business, machine learning (ML) is well established as an essential element in many operations, from finance and logistics to marketing and strategy. However, ML as an analytical tool is still far from widespread in international business (IB) as a science. In this article, we offer arguments as to why this should change by providing illustrative analyses with simulated and real data. We argue that IB as a research community could produce substantial progress if algorithmic ML techniques were adopted as part of the standard analytical toolkit, next to traditional probabilistic statistics. This is not only so because ML improves predictive accuracy but also because doing so would permit empirically addressing complexity and facilitate theory development in IB that does justice to the complex world of international businesses. Along the way, we provide tips and tricks by way of practical tutorial, all relating to a typical ML process pipeline.

Plain language summary

In the research realm, scientists and scholars often aim to comprehend complex occurrences by pinpointing patterns and connections within data. Conventional statistical methods, though potent, may occasionally be inadequate when dealing with the complex, non-linear interactions that characterize real-world data. This is when machine learning (ML – a branch of artificial intelligence that automates data analysis, enabling computers to learn and adapt through experience) can be employed, providing a toolkit that can manage complexity without being overwhelmed by it. The study under discussion connects conventional statistics and ML, demonstrating how ML can aid in theory development, particularly in areas like international business. The research includes an exploratory analysis, utilizing a dataset involving Dutch SMEs (small and medium-sized enterprises), to exhibit ML's potential. The researchers employed various ML techniques, such as Bayesian additive regression trees (BART), to forecast outcomes like the share of revenue earned internationally by these SMEs. Their tactic was to let the data direct the analysis, instead of imposing preconceived models, enabling a more genuine discovery of underlying patterns. The research results were remarkable. ML algorithms surpassed conventional linear models, even when the data was not excessively complex. This confirmed that ML could effectively manage both large and small datasets. The primary discoveries included identifying non-linear relationships and interactions between variables that traditional methods might overlook. Furthermore, the study revealed that ML could provide insights into the significance of different variables and their impacts on the outcomes. The researchers concluded that ML should not only be viewed as a prediction tool but also as a method for inductive exploration of intricate patterns in data. This could lead to a better comprehension of phenomena in multiple fields, including IB. They suggested that by incorporating ML into the research process, scholars could begin to formulate and test hypotheses that encompass a higher level of complexity, more accurately reflecting real-world scenarios. This research's potential impact is immense. It implies that ML can aid in revealing the rich interplay within data that traditional methods might miss. For the future, this means that researchers can approach their work with a new toolkit that is aptly suited for modern data sets' challenges. As ML becomes more accessible and understood, it could transform how we conduct research across disciplines, leading to more sophisticated and comprehensive theories that better reflect the world's complexity around us. This text was initially drafted using artificial intelligence, then reviewed by the author(s) to ensure accuracy.

Résumé

Dans le monde réel des affaires internationales (International Business–IB), l'apprentissage automatique (Machine Learning–ML) est bien établi comme un élément essentiel dans de nombreuses opérations, de la finance et de la logistique au marketing et à la stratégie. Cependant, le ML en qualité d’outil analytique est encore loin d'être répandu dans le domaine de l’IB en tant que science. Dans cet article, nous argumentons pourquoi cela devrait changer en apportant des analyses illustratives avec des données simulées et réelles. Nous argumentons que l'IB en tant que communauté de recherche pourrait réaliser des progrès substantiels si les techniques algorithmiques de ML étaient adoptées comme partie intégrante de la boîte à outils analytiques standard, à côté des statistiques probabilistes traditionnelles. Et ceci non seulement parce que le ML améliore la précision des prévisions, mais aussi parce qu’adopter le ML permettrait de traiter empiriquement la complexité et de faciliter le développement des théories en IB qui tient compte du monde complexe des entreprises internationales. En cours de route, nous élaborons des conseils et des astuces par le biais d'un tutoriel pratique, tous liés à un pipeline de processus de ML typique.

Resumen

En el mundo real de los negocios internacionales, el aprendizaje automático de las maquinas por sí mismas (machine learning o ML en inglés) está bien establecido como elemento esencial en muchas operaciones, de las finanzas a la logística al marketing y la estrategia. Sin embargo, el aprendizaje automático de las maquinas como una herramienta analítica esta aún muy lejos de ser ampliamente extendidos en Negocios Internacionales (IB por sus iniciales en inglés) como ciencia. En este artículo, ofrecemos argumentos sobre como esto debe cambiar, proporcionando análisis ilustrativos con datos simulados y reales. Argumentamos que la comunidad investigadora de Negocios Internacionales podría progresar sustancialmente si las técnicas algorítmicas de aprendizaje automático se adoptaran como parte del conjunto de herramientas analíticas estándar, junto a la estadística probabilística tradicional. Esto no sólo se debe a que el aprendizaje automático mejora la precisión predictiva, sino también a que permitiría abordar empíricamente la complejidad y facilitar el desarrollo de teorías en Negocios Internacionales que hagan justicia al complejo mundo de los negocios internacionales. A lo largo del proceso, ofrecemos algunos trucos y artimañas a modo de tutorial práctico, todos ellos relacionados con un proceso típico del aprendizaje automático.

Resumo

No mundo real dos negócios internacionais, o aprendizado de máquina (ML) está bem estabelecido como um elemento essencial em muitas operações, desde finanças e logística a marketing e estratégia. No entanto, ML como ferramenta analítica ainda está longe de estar difundida em Negócios Internacionais (IB) como uma ciência. Neste artigo, oferecemos argumentos sobre porque isso deveria mudar, fornecendo análises ilustrativas com dados de simulação e reais. Argumentamos que IB, como comunidade de pesquisa, poderia produzir um progresso substancial se técnicas algorítmicas de ML fossem adotadas como parte do kit de ferramentas analíticas padrão, cerca das estatísticas probabilísticas tradicionais. Isto não acontece apenas porque ML melhora a precisão de previsões, mas também porque fazê-lo permitiria abordar empiricamente a complexidade e facilitaria o desenvolvimento de teorias em IB que façam justiça ao mundo complexo dos negócios internacionais. Ao longo do caminho, fornecemos dicas e truques por meio de tutoriais práticos, todos relacionados a um pipeline típico de processo de ML.

摘要

在国际商业的现实世界中, 机器学习 (ML) 已成为从金融和物流到营销和战略等许多运营中的基本要素。然而, ML作为一种分析工具在作为一门科学的国际商务 (IB) 中还远未得到普及。在本文中, 我们通过提供模拟数据和真实数据的说明性分析来论证为什么这种情况应该改变。我们认为, 如果机器学习算法技术被采用作为仅次于传统概率统计标准分析工具包的一部分, IB 作为一个研究社区可以取得实质性的进展。这不仅是因为机器学习提高了预测的准确性, 还因为这样做会允许在实证上解决复杂性并促进 IB 的理论发展, 从而公平地对待复杂的国际商业世界。一路上, 我们通过实用教程提供提示和技巧, 所有这些都与典型的 ML 过程管道相关。