Adaptive Denoising of Sequential Data with Multi-Objective Genetic Algorithms

Redundant or questionable states in sequential data can significantly undermine the reliability of data mining. An effective denoising process should adaptively filter these noisy states while preserving essential information, avoiding critical data loss (over-sanitization). The complexity of sequential data poses additional denoising challenges, yet research on adaptive denoising techniques is limited. We propose a novel MOEA-based approach that uses sequential attributes to guide denoising. By selectively incorporating states within sequences, our approach adapts fitness functions to filter noise. Our approach is validated on students' interactions with an Intelligent Tutoring System, our method effectively balances competing objectives and outperforms traditional denoising strategies.