The sea is important in life on earth, such as regulating the climate, producing oxygen, providing medicines, habitats for marine animals, and feeding millions of people. We must ensure that the sea continues to meet our needs without sacrificing the people of future generations. The sea regulates the planet's climate and is a significant source of nutrients. The sea becomes an important part of global commerce, while the contents of the ocean become the solution of human energy needs today and the future. The wealth and potential of the sea as a source of energy for humans today and the future needs to be mapped and described in order to provide a picture of marine potential to all concerned. As part of the government, the Ministry of Marine Affairs and Fisheries is responsible for the process of formulating, determining, and implementing policies in the field of marine and fisheries based on the results of mapping and extracting information from existing conditions. The results of this information can be used to predict the marine potential in a marine area. This prediction process can be developed using data mining techniques such as applying association rule by looking at the relationship between the quantity of fish based on the plankton abundance index. However, this association rules data mining techniques that requires complete data, which are data sets with no missing values to generate interesting rules for detection systems. The problem often found is that marine data required is not available or marine data is available, but it contains incomplete data. To address this problem, this paper introduces a relative tolerance relationship to the rough set (RTRS). Novelty RTRS differs from previous rough approaches that use tolerance relationships, non-symmetric equation relationships, and limited tolerance relationships. The RTRS approach is based on a limited tolerance relationship taking into account the relative precision between two objects and therefore, this is the first job to use relative precision. In addition, this paper presents the mathematical approach of the RTRS and compares it with other existing approaches using the marine real dataset to classify the marine potential level of the region. The results show that the proposed approach is better than the existing approach in terms of accuracy.