Abstract:
Learning analytics is a data-driven methodology that provides instructors with important
insights regarding student interactions with course materials, allowing them to make
informed decisions about content delivery and structure. We investigated the use of spatial
learning analytics in the context of online and remote education, with a focus on fulfilling
the United Nations Sustainable Development Goal (SDG) 4 - guaranteeing inclusive and
high-quality education for everyone. By combining data from Moodle logs and device
usage and user activity metrics from Microsoft Teams, this study aimed to predict students
quiz performance and enhance personalized learning interventions. The primary objectives
were to create accurate predictive models for student learning outcomes and evaluate their
effectiveness. Additionally, mapping tools were utilized such as Leaflet and ArcGIS to
craft interactive maps, enriching the data-driven learning journey. Study employed
supervised machine learning techniques, including Random Forest, Decision Trees,
Support Vector Machine, LightGBM, K-Nearest Neighbors, Logistic Regression, and
Neural Networks. These predictive models were trained and tested on the preprocessed
dataset to predict quiz scores using binary classification method technique with threshold
of 10. The predictive models' performance was evaluated using metrics such as precision,
recall, F1 score, training time, and prediction time. Results indicate that the Support Vector
Machine (SVM) model achieved the highest recall (100%) and F1 score (85.19%). Logistic
Regression and Neural Networks also performed well, with Logistic Regression showing
a recall of 95.57% and an F1 score of 83.84%, and Neural Networks exhibiting a recall of
94.38% and an F1 score of 83.56%. This study contributes to learning analytics by
demonstrating the potential of spatial data in predicting and improving student outcomes,
aligning with United Nations Sustainable Development Goal 4 for accessible and equitable
education. Future research can refine these models by incorporating additional data sources
and advanced machine learning techniques. Enhancing geographic insights and addressing
ethical considerations in data usage will be crucial.
