Bulletin of Electrical Engineering and Informatics Vol. 14, No. 2, April 2025, pp. 1223~1230 ISSN: 2302-9285, DOI: 10.11591/eei.v14i2.8572  1223 Journal homepage: http://beei.org Enhancing recommendation diversity in e-commerce using Siamese network and cluster-based technique Abderaouf Bahi 1 , Ibtissem Gasmi 1 , Sassi Bentrad 2,3 , Ramzi Khantouchi 1 1 Computer Science and Applied Mathematics Laboratory, Chadli Bendjedid El Tarf University, El Tarf, Algeria 2 LISCO Laboratory, Badji Mokhtar-Annaba University, Annaba, Algeria 3 National Higher School of Cyber Security (NSCS), Mahelma, Algeria Article Info ABSTRACT Article history: Received Apr 8, 2024 Revised Oct 9, 2024 Accepted Oct 17, 2024 This study investigates the difficulty of improving product recommendations in e-commerce systems by tackling the common problem of poor diversity in suggestions. We present a novel approach that uses a Siamese network architecture and ResNet for feature extraction to recommend visually similar elements while incorporating diversity through a cluster-based mechanism. The Siamese network is used to compare product pairs, allowing it to recommend both comparable and dissimilar items from distinct clusters. The model was evaluated using a variety of evaluation metrics, resulting in an accuracy of 88.5%, a precision of 90.2%, a recall of 87.1%, and an F1 score of 88.6%. Our results demonstrate that our strategy maintains a high level of relevance in suggestions while efficiently incorporating variety, hence improving the overall user experience in e-commerce applications. Keywords: Clustering Deep learning Diversity E-commerce Recommender system This is an open access article under the CC BY-SA license. Corresponding Author: Abderaouf Bahi Computer Science and Applied Mathematics Laboratory, Chadli Bendjedid El Tarf University El Tarf 36000, Algeria Email: a.bahi@univ-eltarf.dz 1. INTRODUCTION E-commerce is reliant on recommender systems, which customize user experiences by suggesting pertinent products based on past preferences and behavior [1]-[5]. These systems sometimes implement collaborative filtering, content-based filtering, or a hybrid approach to provide recommendations. However, despite their widespread use, the absence of variation in recommendations is a significant drawback of traditional recommender systems [6]. Researchers refer to this phenomenon as the "filter bubble," in which consumers are perpetually confronted with products that are similar to their own, thereby limiting their capacity to identify novel and diverse products that align with their preferences [7], [8]. Improving user happiness and increasing involvement depends on addressing the diversity in recommendations [9]-[11]. Using computer vision models such as ResNet, which extract rich visual characteristics from product photos [12], [13] to produce more diverse suggestions, one efficient strategy is to include visual variation into product recommendations [14]. Wang et al. [15] developed a deep ranking model to assess fine-grained image similarity in e- commerce, leveraging multiscale networks and an efficient triplet sampling algorithm to surpass traditional methods. He et al. [16] introduction of ResNet marked a significant breakthrough in image recognition by enabling residual learning to train deep networks for improved accuracy and depth while simplifying optimization. The research conducted by Simonyan and Zisserman [17] focused on the important role that network depth has in improving picture identification. This idea was then further developed in the building of VGGNet's incredibly deep convolutional networks. Lecun et al. [18] foundational work pioneered neural