Application of Long-Term Time Series Forecasting in E-Learning

The E-Learning platform plays a crucial role in higher education teaching. Accurately predicting long-term future access patterns based on the platform's log data is essential for optimizing resource allocation in advance. However, many existing methods face challenges in both prediction period and accuracy. To this end, we propose a deep learning-based multi-period modeling network (MPMN) for analyzing log data analysis from the E-Learning platform. It learns the educational cyclical patterns to achieve longer-term and more accurate prediction. By conducting comparative analysis of common time series prediction methods from multiple dimensions, this paper highlights the superiority of the proposed method. The long-term time series forecasting analysis method could effectively assist E-Learning platform administrators in understanding the platform's access load over future periods, enabling them to plan resource allocation in advance, thereby improving teaching quality and enhancing the user experience for both teachers and students.