International Journal of Electrical and Computer Engineering (IJECE) Vol. 13, No. 6, December 2023, pp. 6302~6311 ISSN: 2088-8708, DOI: 10.11591/ijece.v13i6.pp6302-6311  6302 Journal homepage: http://ijece.iaescore.com A brief study on rice diseases recognition and image classification: fusion deep belief network and S-particle swarm optimization algorithm Miryabbelli Jayaram 1 , Gudikandhula Kalpana 2 , Subba Reddy Borra 3 , Battu Durga Bhavani 3 1 Department of Computer Science Engineering (Data Science), Sreyas Institute of Engineering and Technology, Hyderabad, India 2 Department of Computer Science and Engineering (AI and ML), Malla Reddy Engineering College for Women (UGC-Autonomous), Hyderabad, India 3 Department of Information Technology, Malla Reddy Engineering College for Women (UGC-Autonomous), Hyderabad, India Article Info ABSTRACT Article history: Received Nov 6, 2022 Revised Feb 28, 2023 Accepted Mar 9, 2023 In the regions of southern Andhra Pradesh, rice brown spot, rice blast, and rice sheath blight have emerged as the most prevalent diseases. The goal of this research is to increase the precision and effectiveness of disease diagnosis by proposing a framework for the automated recognition and classification of rice diseases. Therefore, this work proposes a hybrid approach with multiple stages. Initially, the region of interest (ROI) is extracted from the dataset and test images. Then, the multiple features are extracted, such as color-moment- based features, grey-level cooccurrence matrix (GLCM)-based texture, and shape features. Then, the S-particle swarm optimization (SPSO) model selects the best features from the extracted features. Moreover, the deep belief network (DBN) model trained by SPSO is based on optimal features, which classify the different types of rice diseases. The SPSO algorithm also optimized the losses generated in the DBN model. The suggested model achieves a hit rate of 94.85% and an accuracy of 97.48% with the 10-fold cross-validation approach. The traditional machine learning (ML) model is significantly less accurate than the area under the receiver operating characteristic curve (AUC), which has an accuracy of 97.48%. Keywords: Deep belief networks Image classification Image recognition Rice diseases S-particle swam optimization algorithm This is an open access article under the CC BY-SA license. Corresponding Author: Miryabbelli Jayaram Department of Computer Science Engineering (Data Science), Sreyas Institute of Engineering and Technology Hyderabad, India Email: drjayaram2022@gmail.com 1. INTRODUCTION Among the most significant food crops for our population is rice. The production of the rice planting business has been rising steadily in recent years, but there are still certain negative elements, including rice diseases, that are reducing its yield. In the frigid regions of northern China, rice brown spots, sheath blight, and blast are currently the most prevalent diseases [1], [2]. It can happen at any time during the rice plant’s growth cycle, which has a significant influence on the rice’s quality and yield and, in extreme situations, can result in no output. This will not only have a negative influence on farmer income but also negatively impact China ’s fiscal revenue and food security. Now, the methods for identifying rice diseases primarily rely on farmers’ judgment, checking disease books or doing an Internet search, consulting agricultural technicians, or asking a plant expert for assistance. Human eyes are not very good at identification, and recognition is much more individualized. A lapse in judgement prevents the proper identification of obstacles to achieving diseases. The