Nanyang Technological University, Singapore: Supporting At-Risk Students With AI

Nanyang Technological University, Singapore (NTU Singapore), one of Singapore’s premier educational institutions, is dedicated to enhancing student outcomes through innovative teaching methods and technology. The Early AleRt for Learning Intervention (EARLI) project is a data-driven initiative aimed at identifying at-risk students early in their academic journey, ensuring timely interventions, and improving overall student success.

With the growing complexity of managing large datasets and applying machine learning (ML) models across multiple disciplines, NTU faced challenges in efficiently processing and analyzing data for academic support.

The EARLI Project

The EARLI project’s primary objective was to predict which second- and third-year students might need academic intervention before the semester begins. However, several challenges arose:

• Manual Data Processing: The traditional approach relied heavily on manual data extraction and processing, which was both time-consuming and prone to delays.
• Inefficient Model Management: Retraining models each semester and monitoring their performance required significant time and effort from the team.
• Scalability: The need for a streamlined and scalable approach to handle data from multiple schools and courses across the university was becoming increasingly difficult.

To address these challenges, NTU implemented Dataiku to enable the automation of their data processing pipeline and streamlined the predictive modeling efforts:

We are the first among the Singapore universities to use Dataiku to address the challenges and needs in a sustainable way. As a comprehensive data science platform, Dataiku enabled us to streamline the process and enhance the uniformity of model performance across semesters. Wei Qiu Research Fellow at Nanyang Technological University

1. Automated Data Pipeline:
Dataiku’s integration with Snowflake allowed seamless data extraction, cleaning, and preparation. Before Dataiku, the IT department was responsible for data extraction. Now, the automation of the process eliminated the need for manual data handling, enabling a smooth and continuous flow of information, as well as less reliance on IT teams.

The integration with Snowflake ensured a seamless data flow, while Dataiku's transparency and explainability tools enhanced governance, facilitating model performance tracking. Wei Qiu Research Fellow at Nanyang Technological University

2. ML Models:

NTU employed AutoML capabilities in Dataiku to develop ML models for predicting at-risk students. This involved four different models based on historical student data, with each model trained for specific courses. The Dataiku platform enabled the creation, training, and evaluation of these models, enhancing the precision and reliability of predictions.

3. Scalable and Consistent Monitoring:
The automation capabilities within Dataiku allowed NTU to scale their processes across multiple schools and courses while ensuring consistent model performance across semesters. By automating the retraining of models each semester, the university could quickly adapt to new data and predict at-risk students accurately.

4. Integration With Qlik Cloud:
Post-processing results were fed into Qlik Cloud, providing real-time access to the insights via interactive dashboards. This allowed care managers to quickly identify at-risk students and provide the necessary interventions.

The user-friendly interface and the seamless integration with various machine learning frameworks were key factors in our decision-making process. The solution provided comprehensive automation, management, and monitoring capabilities for all steps within our data processing pipeline, from data preparation to post-processing. Wei Qiu Research Fellow at Nanyang Technological University

Improvements in NTU’s Daily Operations

By leveraging advanced ML and automation capabilities in Dataiku, NTU’s EARLI project has successfully transformed its approach to identifying at-risk students. The ability to scale operations, automate workflows, and integrate with other platforms has not only improved the efficiency of academic interventions but also enhanced the overall student experience.

Efficiency Gains

Automating data extraction, cleaning, and model training has freed up valuable time for the team to focus on refining models and implementing new predictive approaches.

Seamless Collaboration for Faster Decisions

With real-time data insights provided via Qlik Cloud, the university can make quicker decisions regarding academic support. Dataiku’s user-friendly interface also enabled better collaboration between departments, enhancing the workflow between data science and academic support teams.

A Tangible Impact

The EARLI project achieved a true positive rate of over 70%, meaning that the majority of identified at-risk students genuinely required intervention. By identifying and supporting at-risk students early, NTU has contributed to improved retention rates and better overall academic outcomes. Interventions based on data insights resulted in improved academic performance for at-risk students.

Dataiku has delivered significant value by enhancing team efficiency and agility. By leveraging the automation capabilities of the platform, we have been able to streamline our entire pipeline, eliminating manual tasks and reducing our reliance on the IT department. This allowed us to process data and retrain models more rapidly, thereby ensuring the provision of timely support to at-risk students. Wei Qiu Research Fellow at Nanyang Technological University