PMCT: Parallel Multiscale Convolutional Temporal model for MOOC dropout prediction

Ke Niu

doi:10.1016/j.compeleceng.2023.108989

PMCT: Parallel Multiscale Convolutional Temporal model for MOOC dropout prediction

2023-12-01

1

PRE

AI

DOI：10.1016/j.compeleceng.2023.108989

Original

Original request for help

Save

Abstract

En 中文

High dropout rates are a major challenge for Massive Open Online Courses (MOOCs), making dropout prediction a crucial task. Existing research often develops various prediction models based on student behavioral features. However, most of the existing methods rely on single-scale convolution operations, which can only extract single time scale information. This approach has limited ability in extracting multi-scale features and global information of student behavior, and cannot take into account student groups with different behavior patterns at the same time. To solve this limitation, we propose the Parallel Multiscale Convolutional Temporal model (PMCT) for modeling and predicting student dropout behavior. Specifically, our model uses a Parallel Multiscale Convolution module (PMC) to extract multiscale behavioral features in parallel based on the different features displayed by students at different time scales. Additionally, we employ a stacked residual block of Temporal Convolutional Network (TCN) to extract the potential correlation between scales from the behavior features and as a supplement to the timing information. Experimental results on two large datasets demonstrate that our proposed model outperforms the baseline methods in terms of predictive performance.

Keywords:

Massive Open Online Courses

Dropout prediction

Parallel Multiscale Convolution

Temporal Convolutional Network

PMCT: Parallel Multiscale Convolutional Temporal model for MOOC dropout prediction

Abstract

Journal

Organization