Evaluating and selecting suitable suppliers involves several criteria, making it a complex and difficult problem for decision-makers. Several approaches have been developed to help achieve the best possible results for selecting the most suitable suppliers. Traditionally, multi-criteria decision-making methods, such as the recently introduced MARCOS (Measurement of Alternatives and Ranking according to COmpromise Solution) method, have been used for supplier evaluation and selection. A common issue to be addressed in such methods is determining appropriate weights for the evaluation criteria. We propose an integrated approach that combines machine learning with the MARCOS method for this evaluation process. The approach uses feature importance methods based on interpretable tree-based machine learning to calculate weights for supplier selection criteria. These weights are then used as part of the MARCOS method to rank individual suppliers and select the supplier with the highest rank. The proposed approach is validated through a real-world case study in the oil and gas sector and is compared with state-of-the-art approaches in multi-criteria decision-making and machine learning-based classification. Results show that the proposed model selects suppliers in line with human evaluators and performs similarly to other multi-criteria decision-making approaches. The proposed approach can be useful for designing effective and efficient supplier evaluation systems.