Online supplementary education has been prevalent in recent years due to the advent of technology (e.g., live streaming) and the COVID-19 pandemic. However, the performance of students in this mode of education varies greatly, and the underlying reasons are yet to be investigated. This study aims to understand the impact of various factors and giving their quantified importance, including student information, family information, and course information by applying Machine Learning method to one of the world's largest online learning platforms. Big data analysis is employed for this purpose via leveraging the abundance of data generated from online education platforms, providing insights that are not attainable through conventional methods such as panel surveys or questionnaires used in offline education. The findings indicate that the most significant factor affecting student performance is the disparity in access to educational resources and the socioeconomic status of families. Finally, we predict students' online learning performance by using explainable machine learning with different groups of features. Our results indicate approximately 70% accuracy in predicting students' performance progress.