The procurement process is crucial for project success, but selecting the right supplier is challenging due to risks like delays, cost overruns, and quality issues. Traditional methods often fail to handle these complexities effectively. Data-driven approaches are necessary to improve procurement decisions. This research proposes a Big Data-based framework to optimize supplier selection by evaluating risks and costs. The framework uses the ProZorro Ukraine public procurement dataset, including historical tender data, supplier performance, and market trends. A TabTransformer model processes both categorical and numerical data using advanced deep learning techniques. The model predicts supplier risk and optimizes costs more accurately than conventional methods. Python was used for implementation, and performance metrics RMSE (0.0007), MAE (0.0067), and MAPE (0.1000) confirm high accuracy. Results demonstrate that the proposed model outperforms traditional procedures in supplier evaluation.
Zhang et al. (Wed,) studied this question.