Educational data mining (EDM) is the field of data mining specialized educational data. To get higher learning effect using an intelligent tutoring system, such as e-learning system, it is necessary to grasp the higher accuracy student knowledge state. The purpose of student modeling is the assessment of students’ skills from log data such as examination results, and estimation whether a student solve a question or not. In this research, we propose a student skill modeling method using factorization machines. Factorization machines is a combination of the advantages of support vector machines (SVM) with factorization models. The results of conventional methods which predict student performance using factorization machines showed better than before ones. Especially, we use convex factorization machines with improved factorization machines to predict student skill state in order to improve student modeling. The previous research has predicted whether students would answer the next question or not. Our proposed method predicts whether students have the skill to solve the next question or not.