A survival analysis based volatility and sparsity modeling network for student dropout prediction

Abstract

Student Dropout Prediction (SDP) is pivotal in mitigating withdrawals in Massive Open Online Courses. Previous studies generally modeled the SDP problem as a binary classification task, providing a single prediction outcome. Accordingly, some attempts introduce survival analysis methods to achieve continuous and consistent predictions over time. However, the volatility and sparsity of data always weaken the models’ performance. Prevailing solutions rely heavily on data pre-processing independent of predictive models, which are labor-intensive and may contaminate authentic data. This paper proposes a Survival Analysis based Volatility and Sparsity Modeling Network (SAVSNet) to address these issues in an end-to-end deep learning framework. Specifically, SAVSNet smooths the volatile time series by convolution network while preserving the original data information using Long-Short Term Memory Network (LSTM). Furthermore, we propose a Time-Missing-Aware LSTM unit to mitigate the impact of data sparsity by integrating informative missingness patterns into the model. A survival analysis loss function is adopted for parameter estimation, and the model outputs monotonically decreasing survival probabilities. In the experiments, we compare the proposed method with state-of-the-art methods in two real-world MOOC datasets, and the experiment results show the effectiveness of our proposed model.

Deep Survival Analysis With Clinical Variables for COVID-19

“…In Table 4 , we compared our 1D CNN model’s results with those of other 1D CNN architectures for classifying short and long survival groups. Our findings suggest that our model’s performance is comparable to state-of-the-art models, as described in [45] , [46] , [47] , and [48] . Specifically, we achieved a higher AUC (90.25% versus 84.36–88.10%) and accuracy (83.75% versus 79.06–81.94%) than these previous CNN architectures.…”

“…For a censored patient (i.e., if the person was alive at the end of the study or was lost to follow-up at any time during this study), the survival imputation technique was used for IV, we compared our 1D CNN model's results with those of other 1D CNN architectures for classifying short and long survival groups. Our findings suggest that our model's performance is comparable to stateof-the-art models, as described in [45]- [48]. Specifically, we achieved a higher AUC (90.25% versus 84.36-88.10%) and accuracy (83.75% versus 79.06-81.94%) than these previous CNN architectures.…”

“… Heatmap of significance value ( \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} }{}$-\log _{10} (\textrm {p-value})$ \end{document} ) from chi-square test to measure the differences between two CNN model of the predicted groups (e.g., predicted groups using Convolution-LSTM [45] versus SAVSNet [48] ), green-black circle represents the significance difference between two predictive models in predicting the shorter-term from longer-term survival with \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} }{}$\text{p}\,< \,0.05$ \end{document} . …”

“…Fig. 5: Heatmap of significance value (− log 10 (p-value)) from chi-square test to measure the differences between two CNN model of the predicted groups (e.g., predicted groups using Convolution-LSTM [45] versus SAVSNet [48]), green-black circle represents the significance difference between two predictive models in predicting the shorter-term from longer-term survival with p < 0.05. groups with p-value of 4.1×10 −8 and 2.7×10 −8 , respectively. We note that the convolutional layers (filters) extract features that were used to adjust their weights.…”

Extending Bayesian Personalized Ranking with Survival Analysis for MOOC Recommendation

“…Wintermute et al [19] modeled the certificate rates of MOOC users with a Weibull survival function, following the intuition that students "survive" in a course for a particular time before stochastically dropping out. Pan et al [15] proposed a more sophisticated SA deep learning approach to address volatility and sparsity of the data, that moderately outperformed Cox. However, to the best of our knowledge, such time to dropout has never been incorporated in MOOC recommendations.…”