Balancing Fined-Tuned Machine Learning Models Between Continuous and Discrete Variables - A Comprehensive Analysis Using Educational Data

Article Education & Educational Research

MOOC student dropout prediction model based on learning behavior features and parameter optimization

Cong Jin

Summary: This paper proposes a method to predict students' dropout status using their learning behavior data. By designing a feature extraction method and an intelligently optimized SVR model, this method outperforms other benchmark models in predictive performance.

INTERACTIVE LEARNING ENVIRONMENTS (2023)

Add to Collection

Proceedings Paper Computer Science, Artificial Intelligence

Capturing Fairness and Uncertainty in Student Dropout Prediction - A Comparison Study

Efthyvoulos Drousiotis et al.

Summary: This study aims to explore and improve ways of handling continuous variable dataset to predict student dropout in MOOCs, by implementing various models including RNN and tree-based algorithms. The research shows that using discrete variable models after converting time-series data through feature engineering outperforms time-series models, emphasizing the importance of handling uncertainty in data.

ARTIFICIAL INTELLIGENCE IN EDUCATION (AIED 2021), PT II (2021)

Add to Collection

Proceedings Paper Computer Science, Artificial Intelligence

Early Predictor for Student Success Based on Behavioural and Demographical Indicators

Efthyvoulos Drousiotis et al.

Summary: The study aims to explore and implement an efficient predictive model using Open University Learning Analytics Dataset, to predict students' performance in the course. The model, implemented through machine learning methods, accurately predicts the final result of students at the beginning of the course, helping to reduce dropout rates.

INTELLIGENT TUTORING SYSTEMS (ITS 2021) (2021)

Add to Collection

Article Computer Science, Theory & Methods

A survey on data-efficient algorithms in big data era

Amina Adadi

Summary: This work investigates the issue of algorithms' data hungriness in Machine Learning, providing a comprehensive review of existing data-efficient methods categorized into four groups. These methods include using unsupervised algorithms, creating more data, transferring knowledge between domains, and modifying data-hungry algorithms. The interplay between these strategies is emphasized to explore more robust and data-efficient algorithms, while also discussing limitations, research challenges, and future opportunities in advancing data efficiency in machine learning.

JOURNAL OF BIG DATA (2021)

Add to Collection

Article Education & Educational Research

Predictive learning analytics using deep learning model in MOOCs' courses videos

Ahmed Ali Mubarak et al.

Summary: Analysis of learning behavior of MOOC enthusiasts, particularly related to video lecture data, helps explore different learning patterns and predict learners' performance. This paper utilizes LSTM on video clickstream data to predict weekly performance with an accuracy rate of 93% surpassing other models.

EDUCATION AND INFORMATION TECHNOLOGIES (2021)

Add to Collection

Proceedings Paper Computer Science, Artificial Intelligence