Bananas are one of the major fruit crops that can bring a wealth of income to farmers and the country's economy. But farmers do not fully exploit this potential of banana harvesting because there are various threats to reducing production. One of the biggest threats to banana cultivation is reduced yields due to pests and diseases, and serious economic losses for farmers. Early detection and identification of banana leaf diseases allows farmers to better manage the severity of the disease. There are many symptoms of these pests and diseases. In Some banana crop diseases appear at an early stage while others develop only late because there is no way to save the banana crop. However, banana leaves are highly exposed to diseases such as black sigatoka, yellow sigatoka, bunchy top, panama wilt, streak virus. This thesis proposed an approach that automatically classifies banana leaf disease from an image based on the conventional neural network. Nowadays, banana crops have become much more important than they used to some years ago where they have been only used to feed mankind as well as animals. Banana leaf diseases are currently classified using methods that require a lot of manual work with experts, agricultural extension workers and farmers which are both time-consuming and error-prone. To automate the process of classifying banana leaf diseases, various researchers have found several methods using both machine learning and image processing. So we proposed to create a model that distinguishes banana leaf disease by machine learning (ML) techniques with deep learning classifier like Gray level co-occurrence matrix(GLCM) for feature extraction, k means clustering for segmentation, convolutional neural networks (CNN) for the classification of banana leaf disease through leaves images of healthy or diseased leaf. However, these proposed methods still have limitations. The steps followed in this research for classifying the banana leaf disease are dataset collection, image pre-processing, segmentation, feature extraction, and classification. The model is trained using 615 sample images of banana leaves collected from Arba Minch banana crop farms and other image repository. The training data is randomly split into 80% training and 20% testi ng. The performance of the banana leaf disease classification model achieved an accuracy of 91.41% using the CNN model, and by using texture feature for class’s healthy, yellow Sigatoka, and Panama wilt confusion matrix accuracy was 82.3%, 70.7%, and 63.5% respectively. From the analysis of the experimental results the proposed approach gives the best result. This is due to the fact that convolutional neural network extract high-level features from the input raw data, making it more efficient, accurate, and avoid errors due to subjective manual feature extraction. Keywords: Machine learning technique, deep learning, k means clustering, Convolutional neural networks, and Gray level co-occurrence matrix.