Improving Students' Academic Performance with AI and Semantic Technologies
Project Overview
The document explores the transformative role of generative AI and semantic technologies in education, emphasizing their applications in enhancing student academic performance through dropout prediction and curriculum analysis. It highlights the implementation of advanced machine learning techniques, particularly Long Short-Term Memory (LSTM) networks and Support Vector Machine (SVM)-based Genetic Algorithms, for accurately predicting student dropouts. Additionally, the use of BERT and semantic similarity measurements aids in analyzing course content and identifying course prerequisites, which fosters better curriculum management. Research findings indicate notable improvements in dropout prediction accuracy, showcasing the potential of these technologies to enhance student retention and optimize educational programs in higher education. Overall, the integration of AI in educational settings signifies a substantial advancement towards personalized learning and proactive intervention strategies, ultimately aiming to boost academic success and reduce dropout rates.
Key Applications
Student Retention and Curriculum Analysis using LSTM and BERT
Context: Higher education institutions focusing on student retention, dropout prediction, and curriculum design for course sequencing and similarity measurement.
Implementation: Utilized a combination of Long Short-Term Memory (LSTM) networks for dropout prediction and BERT for semantic analysis of course descriptions. Employed Genetic Algorithm for feature selection and SVM for improved prediction accuracy.
Outcomes: Achieved improved dropout prediction accuracy, enhancing student retention efforts and aiding academic advisors in curriculum development through the identification of course similarities and prerequisite relationships.
Challenges: Dataset imbalance, potential bias in feature selection, data quality issues, and the complexity of integrating AI tools into existing educational frameworks.
Implementation Barriers
Technical
Imbalance in datasets used for training models affecting prediction accuracy, and challenges in data quality and integration of AI tools into existing systems.
Proposed Solutions: Implementing techniques like SMOTE to balance training datasets, focusing on data cleaning, validation, and developing user-friendly interfaces for educators.
Conceptual
Limitations in existing technologies for accurately measuring course similarities and ambiguities in course descriptions.
Proposed Solutions: Further research and development of more sophisticated semantic analysis tools.
Project Team
Yixin Cheng
Researcher
Contact Information
For information about the paper, please contact the authors.
Authors: Yixin Cheng
Source Publication: View Original PaperLink opens in a new window
Project Contact: Dr. Jianhua Yang
LLM Model Version: gpt-4o-mini-2024-07-18
Analysis Provider: Openai