Satyendra Singh Rawat
Department of Computer Science and Engineering, Amity University, Gwalior, Madhya Pradesh, India.

Vikas Thada
Department of Computer Science and Engineering, Amity University, Gwalior, Madhya Pradesh, India.

Amit Kumar Mishra
School of Computer Science and Engineering, Chandigarh University Uttar Pradesh, Unnao, Uttar Pradesh, India.

DOI https://doi.org/10.33889/IJMEMS.2026.11.4.073

  ;

Abstract

Multiclass imbalanced classification is much harder than binary classification because of the major challenges with data sets, including high dimensionality, class overlap, noise, small disjuncts, and a small number of minority class samples. The oversampling technique is largely used to address class imbalance issues. Several oversampling approaches such as SMOTE, ADASYN, Borderline-SMOTE, etc., heavily rely on the k-nearest neighbors (KNN) algorithm, which classifies samples based on the majority class of their selected k-nearest neighbors. Additionally, the Euclidean distance measure imposes further restrictions on the classification process. Due to these concerns, we propose a new hybrid approach called KFCGAN-OVO, which is specifically made for multiclass imbalanced problems, and it does not rely on SMOTE, k-nearest neighbors' selection, or Euclidean distance. The proposed approach employs a Kalman filter for noise reduction and data smoothing; a Conditional Generative Adversarial Network (CGAN), a generative model that generates realistic synthetic data based on given conditions or labels, for dataset balancing; and a one-vs-one (OVO) data decomposition technique to manage high dimensionality. KFCGAN-OVO's effectiveness is analyzed on several measures of data complexity, performance, and computational costs. The results indicate that it outperformed its competitors on all measures.

Keywords- Multiclass imbalance, Data complexity, Generative adversarial network, Kalman filter, OVO decomposition.

Citation

Rawat, S. S. Thada, V., & Mishra, A. K (2026). KFCGAN-OVO: A Novel Kalman Filter-Integrated CGAN Approach for Imbalanced Multiclass Classification. International Journal of Mathematical, Engineering and Management Sciences, 11(4), 1792-1829. https://doi.org/10.33889/IJMEMS.2026.11.4.073.