This study explores the classification of lime diseases using advanced machine learning models and spectrometry data, aiming to improve the accuracy and reliability of disease diagnostics in agriculture. Lime diseases such as citrus canker and citrus black spot cause significant economic losses and challenge traditional visual inspection methods due to their complexity and rapid spread. We utilized Long Short-Term Memory (LSTM), Recurrent Neural Network (RNN), and Bidirectional LSTM (BiLSTM) networks to analyze sequential spectrometry data. These models were further combined with various classifiers, including SVM, KNN, and XGBoost, to enhance the precision of disease detection. The LSTM model alone achieved a precision of 99.96%, while the RNN and BiLSTM models reached 99.4113% and 99.9206%, respectively. Integrating BiLSTM features with classifiers resulted in even higher precision, up to 99.9816%. Our models also excelled in sensitivity and specificity, with the BiLSTM model achieving 99.96% sensitivity and 99.78% specificity. These metrics substantially surpass the results from previous studies, which reported 90.5% sensitivity and 87.0% specificity. This improvement highlights the superior ability of our approach to accurately identify and classify lime diseases. Overall, our research demonstrates that combining deep learning models with traditional classifiers significantly enhances the precision and efficiency of lime disease diagnostics. This advancement is crucial for effective disease management, leading to improved crop protection and economic resilience in the agricultural sector.