Enhancing Early Alzheimer’s Detection Through Machine Learning and Data Preparation
As part of the Curiosity Cup 2025, Team Data ACES from the University of Santo Tomas, the Philippines, tackled the great challenge in modern healthcare: early detection of Alzheimer’s disease.
Their work was awarded as a winner of the Data Preparation Category.
They developed predictive models that can help identify Alzheimer’s in its early stages, when interventions are most effective. The team worked with three open-source datasets from Kaggle, each from a different time frame (2019, 2023, 2024), to compare how data quality and structure affect prediction outcomes.
They used SAS Viya Workbench for Learners, using:
- Gradient Boosting Classifier
- Data cleaning, Feature encoding, and Correlation analysis
- Model validation using accuracy, F1-score, ROC-AUC, and hyperparameter tuning via GridSearchCV
All models showed strong predictive performance, with the best test accuracy reaching 95.8%.
The most important predictors across datasets included:
- MMSE (Mini-Mental State Examination)
- CDR (Clinical Dementia Rating)
- Functional and behavioral assessments
- CSF biomarkers and Age
This project demonstrates how machine learning and the right data tools can support early detection and better management of neurodegenerative diseases.
Well done, the Team Data ACES! Cyrus Canape, Racell Gabriel Sincioco, Edrine Frances Esguerra, Nicolai Jechri Rafael Alcaraz, and the Faculty Advisor Francis Noel Alarcon.
Access to the full paper is available under the article's headline as a PDF.
