Sericulture plays a vital role in India’s agro-industries, with the mulberry plant being essential for cocoon-bearing. The quality and quantity of mulberry leaves are crucial, as they directly impact cocoon production. Therefore, detecting and classifying diseases in mulberry leaves is a necessary process that can assist farmers and boost the economy. To address this need, an innovative Machine Learning (ML) approach is proposed using a Scalable kernel-based Support Vector Machine (S-SVM), which enhances the classification accuracy of traditional Support Vector Machine (SVM) methods. To further optimize the performance of S-SVM, the Al-Biruni Earth Radius Optimization (AERO) technique is introduced, which effectively balances the weight between the majority (healthy) and minority (diseased) classes of the leaves. This approach enables accurate classification of mulberry leaf diseases, such as healthy, leaf spot, and leaf rust. Before classification, image preprocessing is a key step, for which we adopt the K-means clustering algorithm to identify the disease-affected regions on the leaf. Contour tracing is then used to outline these regions for better disease detection. Additionally, Transfer Learning (TL) is applied to extract relevant features from the images, leveraging pre-trained models to improve classification accuracy. The proposed system is simulated using MATLAB software, and its effectiveness is compared with state-of-the-art methods. The results indicated excellent performance, with a training accuracy of 99% and a loss of 0.8%, and a validation accuracy of 98% with a loss of 0.3% over 100 epochs, demonstrating the potential of this approach for real-world application in sericulture.