The purpose of this study is to analyze the main factors influencing the prediction of high teacher efficacy using a random forest (hereafter RF), which is one of the data mining techniques, and to clarify whether these influencing factors differ according to gender, which could provide a basic data to establish a policy that can effectively contribute to the enhancement of teacher efficacy. To do this, this study used the OECD s Teaching and Learning International Survey (2013 survey). The subjects were 1,802 women and 787 men. This study used 311 teacher-level items and 263 principle-level items as explanatory variables, taking advantage of the advantages of RF which is free from multicollinearity problem and the restriction of degrees of freedom. The main results are as follows. First, ‘classroom practices in the target class (teaching and learning activities, students’ problematic behavior, and student learning assessment)’, positive impact of professional development,’ job satisfaction,’ positive change due to feedback were the main factors in teacher efficacy of both male and female teachers. Second, there were similarities and differences between the male and female teachers at the level of the question, even within the factors that were common to male and female teachers. It should be noted that these findings could be found because RF can grasp the effect of the explanatory variable on the item level on the dependent variable. Based on these results, this study suggested policy implications for improving teacher efficacy.