Nonparametric Big Data Analysis and Machine Learning in Supply Chain 4.0

The profound changes brought by Industry 4.0 and its impact on supply chain management, leading to the emergence of Supply Chain 4.0 (SC 4.0). Industry 4.0 is characterized by the integration of advanced technologies such as big data. It represents the fourth industrial revolution and promotes greater automation, connectivity, and real-time data exchange. SC 4.0 leverages these disruptive technologies to streamline supply chain processes, generating significant strategic benefits for stakeholders (Frederico et al. 2020). With digital transformation, supply chains are faced with new challenges to maintain or improve their competitive edge. Unlike in the past, firms now make substantial investments in information technologies, generating large amounts of real-time data (Yu et al. 2018). The rapid development of these technologies has enabled companies to trace the intangible flow of information (Li and Liu 2019). To leverage the full potential of these data sources supply chains need to apply appropriate analytical tools, such as machine learning techniques and big data analysis. However, implementing these tools is not straightforward and demands capabilities in advanced analytics and suitable techniques. This expertise allows supply chains to use all available data effectively (Frazzon et al. 2019), positively impacting decision-making for the supply chain (Lee and Mangalaraj 2022). This thesis introduces new analytical methodologies to leverage the potential of nonparametric big data analysis and machine learning within SC 4.0, emphasizing the critical role of advanced analytics in unlocking the full potential of big data. Following a comprehensive review of the literature, several research questions have been identified, and this thesis aims to address these research problems. To tackle these questions, new methodologies and case studies are illustrated. The thesis introduces three methodologies: extensions of the NonParametric Combination (NPC) method and a new machine learning approach, all designed to tackle the complex data scenarios characteristic of SC 4.0. These methodologies are specifically aimed at addressing key challenges in SC 4.0, including enhancing production processes, supporting product development and improving forecasting accuracy through a more precise and reliable model selection procedure. To demonstrate the effectiveness of these methodologies, several case studies were conducted. The first case study uses an extension of the NPC methodology to optimize the extraction processes of the silk fibroin. The second applies the new machine learning approach to enhance the development of new wireless earphones. Finally, the selection of the most suitable machine learning model for predicting the price and customer satisfaction of new products is addressed using an innovative methodology that combines a permutation-based approach with a robust ranking procedure. In summary, this thesis provides a new perspective on using nonparametric big data analysis and machine learning in SC 4.0. It demonstrates how these advanced analytical tools can unlock meaningful insights, streamline production processes, enhance product development, and identify the most effective machine learning models, offering valuable guidance for practitioners and organizations in the evolving landscape of SC 4.0.