The constructive covering algorithm (CCA) has unique advantages and functions in dealing with multiple-classification problems. To better address the defects and deficiencies of the existing multi-class decision-theoretic rough sets (DTRSs) model, this paper introduces the CCA into DTRSs and constructs a multi-class DTRSs model based on the CCA. First, through the process of machine learning, this method achieves automatic multi-class clustering and effectively addresses the shortcomings of DTRSs in solving multi-class problems, namely, the large amount of computation and the impact of subjective evaluation factors on the inference of threshold values and decision rules as a result of excessive parameters. Second, by optimizing the covering centre and selecting the covering radius, the proposed method improves the efficiency of machine learning. Third, by setting up a distance parameter between covering classes, the method effectively addresses decision redundancy and conflicts when a DTRSs model is used to solve multi-class problems. Then, by organically integrating the cost sensitivity, inter-class distance parameter and class quality function and systematically reclassifying the objects in the boundary domain, it effectively solves the problem of an excessively large boundary domain. Finally, we summarize the decision-making algorithm of the multi-class DTRSs model based on the CCA and demonstrate the feasibility and effectiveness of the method through experiments. INDEX TERMS Constructive covering algorithm, decision-theoretic rough sets, three-way decisions, multiclass.