<p>The identification of phenotypic features for soybean diseases, which significantly limit yield and quality, is crucial for soybean breeding, cultivation, and fine management. Soybeans are an essential source of plant protein and oil. Traditional deep learning models have poor recognition accuracy, and the chemical analysis procedure for soybean illnesses takes a long time. By spotting leaf diseases earlier, farmers may take proactive steps to stop the spread and reduce the impact on crop quality and productivity. As a result, the given paper presents a hybrid framework that consists of a convolutional neural network (CNN) with a long-short term memory (LSTM) model for feature extraction and classification, respectively. The spotted hyena optimizer (SHO) is used to adjust the CNN-LSTM’s weights or hyperparameters. As a set of hyperparameters to be optimized, SHO produces a number of solutions. The procedure is then repeated until the ideal solution is found. The suggested framework outperforms current optimization models and studies due to its adjusted hyperparameters and shorter execution time, which produce higher accuracy (%), precision (%), and recall (%).</p>