Confusion among students hinders learning and contributes to demotivation and disinterest in the course materials. However, it takes a lot of time and resources to identify confused pupils in extensive courses. Using LSTM and Attention, we suggest a deep learning model for monitoring students' confusion by EEG signals from students when they watching MOOC videos. The model obtained an accuracy of 0.82 on the EEG data, exceeding the previous experimental results for this dataset. Experiments show that the attention mechanism picks up on the significance of various features on prediction results. It can effectively solve the overfitting problem and improve the model classification effect.

Cloud significantly impacts the Earth's radiation budget and is also a major source of uncertainty in global climate modeling (Bony et al., 2015;Ramanathan et al., 1989;Zhang et al., 2022). Cloud vertical structure (CVS), characterized by cloud top height (CTH), cloud base height (CBH), cloud thickness and the vertical distribution of multi-layer clouds, significantly impacts the atmospheric circulation by changing the net radiation budget and the energy redistribution at the top of the atmosphere (AT) (Chen et al., 2000;Slingo & Slingo, 1988;Xu et al., 2021). Studying CVS is the important prerequisite to better modeling the cloud feedback process and improving the performance of the general circulation models (Ovchinnikov et al.,

The Tibetan Plateau (TP) is one of the regions with the most remarkable diurnal variations in the global atmosphere. The zonal shear line (ZSL) is the primary synoptic system inducing precipitation over the TP in boreal summer. It is of great significance in studying its diurnal variation. This study objectively identified and composited 11 ZSL cases inducing anomalous heavy precipitation based on the ERA5 hourly reanalysis datasets from June to August during 1980–2019. The diurnal variation characteristics and mechanisms of ZSL’s intensity were explored based on the composited ZSL. Results suggest that the intensity of ZSL has significant diurnal variation, reaching the peak at midnight (about 23 Local Solar Time). The dynamical and thermal structures near the ZSL both show significant diurnal variations. There is a significant heating effect near the ZSL after sunrise, followed by strong divergence (convergence) in the upper (lower) layer and significant increases in water vapor flux convergence and ascending motions. The vorticity is strongest at midnight. Diagnosis by the complete‐form vertical vorticity tendency equation reveals that the thermal effect, dominated by vertically non‐uniform distribution of atmospheric diabatic heating, is the most crucial factor affecting the diurnal variation of ZSL’s intensity. The configuration of solar radiation and sensible heat in the near‐ground layer and latent heat in the middle and upper layers jointly dominate the vertically non‐uniform heating. The horizontally non‐uniform heating only contributes little to the local change of vorticity at 500 hPa. Ascending motion and vorticity advection have only weak effects.