Bilkent University
Department of Computer Engineering

---

CS 590/690 SEMINAR

 

Ensemble-based Regression for Data Streams using MMR

 

Mehmet Kadri Gofralılar
Master Student
(Supervisor: Prof.Dr. Fazlı Can)

Computer Engineering Department
Bilkent University

Abstract: The increasing volume of streaming data requires efficient and accurate predictive models. Traditional machine learning systems often rely on fixed datasets and batch learning, making them unus- able in dynamic environments where data distributions evolve over time. This study introduces Ensemble-based Regression for Data Streams using MMR, which is a novel approach that leverages Max- imal Marginal Relevance (MMR) for real-time numerical estimation tasks. Unlike conventional models, we incorporate ensemble learn- ing strategies where MMR is utilized for component selection. This helps increasing the diversity and relevancy of the models that con- tribute to the final prediction, improving robustness and accuracy over time. Additionally, we propose a dynamic adjustment mech- anism for MMR’s λ parameter, allowing the system to effectively adapt to concept drift. The proposed approach will be evaluated using prequential learning metrics in a real-time data stream envi- ronment, demonstrating its effectiveness in maintaining stable and accurate predictions despite evolving data patterns. With our find- ings, we hope to highlight the potential of adaptive, diversity-aware learning systems in addressing real-time numerical prediction chal- lenges across various domains.

 

DATE: March 24, Monday @ 13:40 Place: EA 502