Agriculture is the foundation of each country on the planet. In India, the greater part of the provincial populace actually relies upon horticulture. The rural area gives significant work in rustic zones. The infections on the banana are a significant issue which makes the sharp decline in the creation of banana. These require cautious determination and ideal dealing with to shield the harvests from weighty misfortunes. Presently a day's yield faces numerous illnesses. The unaided eye perception of specialists is the principle approach embraced by and by for location and distinguishing proof of banana stem illnesses. In any case, this necessities persistent observing of specialists which may be restrictively costly in enormous homesteads. So this paper proposes the classification and feature extraction in detecting the banana stem diseases. When the disease is detected in stem of the plant, it will help the experts in carrying out the prevention methods. Here the enhanced image processing and machine learning techniques are adopted for enhancing the classification accuracy in disease detection. So we use linear principle component analysis (L-PCA) and recurrent neural network (RNN) for extraction and classifying the input images from dataset. We initially pre-process and segment the image, where the segmentation is done using average adaptive thresholding which segment the diseased part based on the threshold value of the image. The simulation result shows the enhanced accuracy, precision and recall.