Building Agile Supply Chains with Supply Chain 4.0: A Data-Driven Approach to Risk Management

The aim of this study is to advance multi-label delivery delay predictions in supply chains using machine learning and deep learning models. The work used Decision Trees, Random Forests, CNN, and FNN on a real-life logistics dataset consisting of customers and products features. EDA and feature selection were examined and performed as a part of the data preprocessing process at the pre-processing step of the models. According to current model results, Random Forest model reached maximum accuracy of 66.5% along with Decision Trees and FNN. CNN, although, worked well in some instances was not up to par in some areas because it overfitted. The results also reveal how Random Forest is a particularly useful algorithm for predicting delivery delays accurately. The conclusion suggests enhancing the deep learning models performance and combining approaches. Further work should also incorporate other variables in order to improve the predictive capability in real-life requirements of supply chain environments including conditions and stocks.