College Student Retention Risk Analysis from Educational Database Using Multi-Task Multi-Modal Neural Fusion

Abstract

We develop a Multimodal Spatiotemporal Neural Fusion network for MTL (MSNF-MTCL) to predict 5 important students' retention risks: future dropout, next semester dropout, type of dropout, duration of dropout and cause of dropout. First, we develop a general purpose multi-modal neural fusion network model MSNF for learning students' academic information representation by fusing spatial and temporal unstructured advising notes with spatiotemporal structured data. MSNF combines a Bidirectional Encoder Representations from Transformers (BERT)-based document embedding framework to represent each advising note, Long-Short Term Memory (LSTM) network to model temporal advising note embeddings, LSTM network to model students' temporal performance variables and students' static demographics altogether. The final fused representation from MSNF has been utilized on a Multi-Task Cascade Learning (MTCL) model towards building MSNF-MTCL for predicting 5 student retention risks. We evaluate MSNF-MTCL on a large educational database consists of 36,445 college students over 18 years period of time that provides promising performances comparing with the nearest state-of-art models. Additionally, we test the fairness of such model given the existence of biases.

University Student Dropout Prediction Using Pretrained Language Models

“…Another study [8] used machine learning techniques to predict freshmen dropout using secondary school academic records and first-year course credits. Several researchers [11,25] predict student dropout using unstructured data. The study [25] extracted student sentiment from advisor notes written by student advisors to predict student dropout.…”

“…The study [25] extracted student sentiment from advisor notes written by student advisors to predict student dropout. In a recent study [11], the authors utilized temporal structured data and unstructured counseling data from all semesters to predict college student dropout using a PLM. However, because these studies train on structured and unstructured data separately, they lack the ability to capture the relationships between the data.…”

“…We evaluated the performance of seven methods. We first adopted multimodal spatiotemporal neural fusion (MSNF) [11] as a state-of-the-art method for the dropout prediction task. Other baselines included machine learning methods, such as logistic regression, support vector machine (SVM), decision tree, random forest, as well as deep learning methods, such as MLP.…”

“…For instance, the GPA and student club participation status are recorded as numerical and categorical data, respectively, whereas course evaluation comments are recorded in textual format. Recently, several studies [10,11] have analyzed unstructured textual data (e.g., advisor notes) to predict student dropout. The study [10] used sentiment analysis to extract positive or negative words from the advisor's notes.…”

“…However, that approach considered only unstructured textual data to predict student dropout. Another study [11] developed a multimodal neural network model that combined structured data (e.g., GPA and SAT scores) and unstructured data (e.g., advising notes and forum posts). The authors showed that fusing structured and unstructured data was effective for predicting student dropout.…”