Journal Article

QRMHF-DNK: Hybrid Optimization and Deep Kernel Approach for Fake News Detection

Devi, K Saritha, Chidambaram, M

2026

Overview

In this study, QRMHF-DNK (Quasi Reflection Metropolis Hasting Firefly- Deep Neural Kernel), a hybrid framework is designed to improve fake news detection on benchmark datasets. The framework integrates three main stages: data preprocessing to reduce sampling errors, feature selection using a swarm-based optimization strategy, and classification using a deep neural kernel model. This combination enables effective handling of high-dimensional textual data while accurately identifying informative features for classification. The proposed framework was evaluated on a publicly available Kaggle fake news dataset and compared with existing cooperative and multilingual deep learning methods. Experimental results show that QRMHF-DNK achieves a precision of 0.98 and recall of 0.95, with a sampling error of 0.0671%, indicating that the sampled data closely represents the true class distribution. These results demonstrate the effectiveness of the proposed approach on the evaluated dataset and suggest its potential applicability to fake news detection tasks, while further validation on additional datasets is left for future work.

• Publisher: CHRIST (Deemed to be University)
• Subject: Datasets