In the field of education, the use of big data to improve students' learning performance has been a very popular topic recently. The attention of students in the curriculum is obviously a key factor affecting learning performance. In past studies, the attention of students is often measured through questionnaires. The results obtained through the questionnaire are not objective enough, and there are problems of those who feeling good or lack of confidence. Therefore, the measurement of attention was gradually changed to a hardware method, such as brain waves or heart rhythm, etc. Although, the hardware measurement method is more accurate, the cost of the equipment greatly increased, and allowing students to wear these hardware devices may cause the students to perform measurements in an unnatural learning environment. As a result, this study attempts to use a software method to measure and analyzes students' attention in class through e-book reading log, and to explore whether attention can be used as an important indicator of predicting learning performance, and design a set of training to improve student's attention. The research results show that both the e-book reading log and the student's question-posing score can be used to measure attention, and student's attention in class is a key factor affecting their learning performance, that is students with high attention tend to achieve better learning performance.