Open Access Article

Title: Unbalanced data identification based on Bayesian optimisation convolutional neural network

Authors: Yanzhen Wang

Addresses: School of Electronic and Computer Engineering, Zhengzhou College of Finance and Economics, Zhengzhou 450044, China

Abstract: The difficulty of unbalanced datasets in classification issues has become more noticeable with the fast expansion of data science and machine learning approaches. When confronted with uneven data, conventional machine learning methods often produce poor prediction of a few classes. Based on Bayesian optimisation (BO), we propose in this work an enhanced convolutional neural network (CNN) framework (BO-CNN) meant to optimise the hyperparameter configuration of CNNs while resolving the class bias problem in unbalanced data. Experimental results reveal that BO-CNN shows benefits on challenging datasets, lowers miss-detection and false alarms, and efficiently enhances the capacity of the model to manage unbalanced data. These results offer a fresh approach for unbalanced data categorisation and a useful guide for the future optimisation and implementation of deep learning models.

Keywords: BO; Bayesian optimisation; convolutional neural network; CNN; unbalanced data recognition.

DOI: 10.1504/IJICT.2025.144056

International Journal of Information and Communication Technology, 2025 Vol.26 No.2, pp.96 - 111

Received: 08 Dec 2024
Accepted: 18 Dec 2024

Published online: 22 Jan 2025 *