DRAGON Module: A plug-and-play encoder that learns global structural patterns using graph attention and diffusion over MdBGs. 
Dual-Branch Architecture: Combines DRAGON with a standard encoder (e.g., CNN) to improve temporal modeling under constrained input
Methods & Key Findings
I am collaborating with Dr. Hasan Kurban (Assistant Professor at Hamad Bin Khalifa University in Qatar), Dr. Mehmet Dalkilic (Professor of Computer Science at Indiana University), and Mert Cakiroglu (PhD student at Indiana University) on a project that uses de Bruijn graphs (dBGs) for time series modeling. dBGs are widely used in computational biology for tasks such as genome assembly and sequence modeling, where they efficiently and compactly represent long sequences from overlapping fragments. Time series forecasting is inherent in many real-world application from health monitoring and finance to industrial system yet many models struggle to capture long-range dependencies and recurring motifs. Applying de Bruijn graphs to time series offers a novel way to capture local transitions, recurring motifs, and global context. It also provides a more interpretable structure, as the graph explicitly encodes common patterns and transitions in the data. Our findings show that representing time series as dBGs can improve forecasting accuracy across several benchmark datasets
Collaborators & Students
- Dr. Hasan Kurban - Assistant Professor at Hamad Bin Khalifa University in Qatar
- Dr. Mehmet Dalkilic - Professor of Computer Science at Indiana University
- Mert Cakiroglu - PhD student at Indiana University
Publications & Presentations
- Cakiroglu, M. O., Altun, I. B., Kurban, H., Khorasani Buxton, E., & Dalkilic, M. (2025, June). Multivariate de Bruijn Graphs: A Symbolic Graph Framework for Time Series Forecasting. In Proceedings of the ICML 2025 Workshop on Foundations of Multimodal Sequential Data (FMSD).
- Cakiroglu, M. O., Kurban, H., Sharma, P., Kulekci, M. O., Buxton, E. K., Raeeszadeh-Sarmazdeh, M., & Dalkilic, M. M. (2024). An extended de Bruijn graph for feature engineering over biological sequential data. Machine Learning: Science and Technology, 5(3), 035020.
- Cakiroglu, M. O., Kurban, H., Buxton, E. K., & Dalkilic, M. (2024, October). A novel discrete time series representation with de bruijn graphs for enhanced forecasting using timesnet. In 2024 IEEE 11th International Conference on Data Science and Advanced Analytics (DSAA) (pp. 1-3). IEEE.