The study aims to investigate the analysis of student metacognition skill in solving the rainbow connection problem under the implementation of the research-based learning model. The method used in this study is a mixed method. The research subjects were the 4th semester of students of higher education consisting of 30 students of experimental class and 45 students of the control class. The instruments of this research are a test, an observation, and an interview. Quantitative method is applied to analyze the difference of student achievement result among two classes, while the qualitative method is applied to analysis the students' metacognitive skills. The results show that there are significant differences between the two classes that applied research-based learning and conventional learning models. The conventional model refers to student-centered learning of Think-Pair- Shares (TPS) model. The statistical result indicates that the (2-tailed) significance of the independent sample t-test in the pre-test was 0.000 or a = 0.05. It implies that the implementation of RBL significantly affects the students' metacognition skill in solving the rainbow connection problem.