Currently, the main work I am doing is to build the predictive models based on machine learning and deep learning algorithms to analyze the students’ online learning behaviors, and the associations between the learning behavior themselves as well as with the learning performances. The goal of the work is to help instructors to have a good understanding about the students’ learning status as early as possible and then provide the most individualized interventions for them to make students have better online learning experience. The specific objectives of the work are as follows:
- Build models that can capture students’ learning behavior patterns in online learning setting accurately;
- Identify the relationships between the currently seen behavior with the latent learning behavior;
- Identify the relationships between the behavior patterns and the learning achievement;
- Identify the learning motivations based on the different learning behavior patterns;
The projects aim to build accurately models for predicting students’ learning performance based on multiple sources of data (e.g., text, audio, video, posture, etc).
The aim of the projects is to apply machine learning and deep learning algorithms to analyze the students mastery levels of the concepts (hidden state that we cannot quantify) involved in the problems they have solved based on the observed performances.
The issues need to pay attention to/working on:
- how to automatically extract the concepts involved in the questions;
- build a model to provide interpretable results in terms of the performances on the currently working on problems based on the mastery levels on the related concepts;
- To provide a way to analyze the levels of different concepts involved in a problem;
- Provide understandable/actionable results for instructors with respect to improve student performances.
- In BKT assumes that all the students have the same initial mastery level for a specific kc.