In light of the COVID-19 pandemic, this research examines supply chain disruptions critically, highlighting their devastating effect on businesses like as drugs and perishable food. The report emphasizes the need of durable supply networks, encouraging firms to invest in strong risk mitigation measures and cutting-edge technology to maintain continuity and adaptation in an ever-changing business environment. Deep learning and machine learning approaches have emerged as critical tools for solving these supply chain difficulties. Decision trees, random forests, k-nearest neighbors, and support vector machines (SVM) have all been shown to be effective in mitigating supply chain issues, with SVM achieving an impressive accuracy rate of 96.6%. Furthermore, the random forest model has a respectable accuracy of 95.00%. Notably, the study discovers that the gated recurrent unit architecture outperforms the long short-term memory design, with an accuracy rate of 98.01%. This investigation provides useful insights for the difficulties confronting contemporary supply networks.
Top1. Introduction
The ongoing COVID-19 outbreak has presented an unparalleled challenge for worldwide supply networks. The worldwide supply chains have been rendered susceptible due to significant impacts or delays in the shipments of essential and life-saving items, including medicines, agriculture, healthcare, and manufacturing (Craighead et al., 2020). A supply chain (SC) refers to a well-coordinated network of people, machinery, activities, resources, and technology that collaboratively engage in the production and distribution of a product to its final consumers. The process involves the transportation of raw materials or partially processed goods from suppliers to manufacturers, as well as the conversion and delivery of the final product or service to end-users or consumers (Bassiouni et al., 2023). The effective management of SC activities is crucial for ensuring the efficient movement of resources and goods.
Supply chains are often complicated and comprise numerous phases, such as manufacturing, packaging, distribution, and retail. Delays and shortages may occur at any stage in the supply chain. Movement limitations, border closures, and a need for transit alternatives (e.g., aircraft and trucks) hampered the movement of products. It influences the delivery of pharmaceuticals and perishable food products. Because supply chains are global in nature, interruptions in one place might result in shortages of vital raw materials or components required for pharmaceutical and food manufacturing (Cundell et al., 2020). Companies experienced difficulties maintaining inventory levels as a result of variable demand patterns and supply interruptions. Overstocking and understocking were challenges that needed to be addressed in real-time.
This study focuses on substantial disruptions in the supply systems for food and medicines. The implementation of lockdown measures, border restrictions, and labor shortages had a detrimental impact on the production, distribution, and transportation of crucial commodities. The occurrence of panic purchasing resulted in the depletion of stocks, hence impacting the availability of essential drugs and basic things. The global health crisis has highlighted the need to implement robust and flexible supply chain strategies. It has led both enterprises and governments to reassess and strengthen their operations in order to guarantee continuous access to essential commodities in times of emergency (Cano-Marin et al., 2023).
Enhanced comprehension of the potential dangers associated with shipments may significantly reduce feelings of anxiety. The subsequent section of the chapter presents many ML methodologies aimed at mitigating the risks associated with loads by forecasting the feasibility of exporting goods from one origin to another, even in light of the limitations imposed by the COVID-19 epidemic. In a similar vein, the work of prediction has also been undertaken using DL methodologies. The ML and DL models under consideration have the potential to analyze the dataset and presented through Figure 1.
Figure 1. Block diagram of machine learning for supply chain disruptions in the COVID-19 pandemic
The challenge included the use of decision trees (DT), random forest (RF) (Ali et al., 2022), k-nearest neighbor (KNN), and SVM algorithms. The approaches used in ML include four primary stages: data acquisition, noise reduction or pre-processing, feature extraction, and classification (Cavalcante et al., 2019). The LSTM and GRU architectures are recurrent neural network (RNN) models that diminish fading gradients and capture long-range connections in sequential data. Continuous data, Time-series data, and related applications benefit from these approaches. The LSTM and GRU designs eliminate vanishing gradients and capture sequential input dependencies. LSTM networks handle long-range relationships better because of their complicated architecture, which includes a memory cell (Jassim et al., 2023). GRUs manage medium-range interactions effectively and are computationally efficient. The choice depends on the task, dataset size, and processing resources (Bi et al., 2022).