This study will explore and experiment with various combinations of methods to handle data imbalance in order to address the common issue of insufficient minority samples in at-risk student prediction. Additionally, we will examine the purpose of applying computer tools to educational issues and emphasize the necessity of adhering to models with high transparency and explainability, ensuring that the decision-making process can be transparent and comprehensive in the context of learning analytics. After comparing model performance, we selected the logistic regression model combined with correlation analysis and threshold adjustment, which showed outstanding performance in UAR, G-means, and other evaluation metrics. We will analyze the reasons behind students' academic performance based on the feature importance ranking from the model, thereby establishing a high-performance and high- transparency benchmark model for the LBLS593 dataset.

Generative AI (GenAl) has demonstrated its benefits for student learning outcomes in various educational settings. Due to its diverse applications, it has become a trend for many researchers to explore its usage. In this study, we applied GenAl, specifically ChatGPT, in a workshop designed for 43 university teachers to understand the importance of Plato's Allegory of the Cave in the context of education. The Allegory was chosen because the university teachers came from diverse disciplines, and we needed a subject area that they were not already familiar with. The workshop utilized GenAl in four main aspects: question creation, topic breakdown, problem-solving, and post-class summaries. After the workshop, we used statistical and natural language processing techniques to analyze the teachers' learning outcomes regarding the Allegory of the Cave and the potential of post-class summaries for assessment. The experimental results showed a significant improvement in the teachers' understanding of Plato's Allegory of the Cave. It also confirmed that post-class summaries, as the highest form of knowledge expression, were significantly related to their learning outcomes.

Nowadays, there are more and more researches focused on prediction of learning outcome, and most of them applied quantitate type of analysis approaches. Thus, we want to apply another type of analysis approach to do early prediction. In this research, we applied temporal features and analysis approach to predict students' learning outcomes and identify at-risk students. The result shows that using temporal features is effective on early prediction of learning outcome and there exists differences of learning behaviors between students which have different learning background.

Many previous studies have shown that students' learning strategies and engagement are both important factors affecting students' learning outcome. We want to know whether the engagement of students who use different learning strategies is different, so that we can proxy measure students' engagement according to the learning strategies used by students. Different from the past methods of using questionnaires or interviews, this research used education data mining technology to directly analyze the learning logs of online learning system to extract students' learning strategies. We collected the learning logs of 56 students, and extracted the learning strategies of the students using sequence clustering. We classified four learning strategies and performed Pearson correlation analysis with students' learning outcome and engagement, and discussed the key strategies that affect students' learning outcome. We also found that students who use different learning strategies have different levels of engagement.

In recent years, educational resources have gradually been digitized, and digital education platforms have gradually become popular. We use AI to accurately assist people in performing daily tasks through a machine learning process. In education, we can use AI in many situations, such as predicting student's learning outcome and discovering student's learning strategies. However, most solutions have not yet utilized modern AI capabilities, such as natural language processing. This research aims to help teachers use machines to automatically generate short answer questions to reduce the time for teachers to write exam questions. In addition, the main reason we focus on short answers is that many studies prove that short answer exercises can enhance student's long-term memory, thereby improving their learning performance. We propose an automatic question generation (AQG) system that combines syntax-base and semantics-base, in order to prove that the system is highly available and improve student's learning performance, we conducted experiments with 41 students. The experimental results show that student's learning performance has been significantly improved, which means that by repeatedly testing the machine question generation system, students can deepen their long-term memory of course knowledge.