Akademik Veri Yönetim Sistemi
International Congress on Information Technologies in Medicine, Pharmacy, Agriculture, Food, Forestry, Environment, and Engineering (INFTEC - 2024), Tokat, Türkiye, 8 - 10 Mayıs 2024, ss.17
Plants play an important role in sustaining life in nature. This is a subject that has attracted human attention for centuries. Therefore, the most important step in human beings understanding of nature is the naming and classification of living things in nature. Classification of plants is the subject of research in various scientific fields, especially in agriculture. Different methods are used for classification. Machine learning is the most used of these methods. For this purpose, logistic regression, one of the methods used in machine learning, is used to perform classification in cases where the dependent variable is categorical. The performance of the logistic regression model used is evaluated with the complexity matrix. This study focused on the classification of leaves. Feature extraction was performed using the edges of the leaves in binary images. Since the edge information obtained from each leaf was not equal, resampling was done using the interpolation approach and classification was done using the logistic regression method. There are a total of 89 leaf images in the dataset, consisting of 8 different leaf types from the literature. 200 different attempts were made to classify these leaves and their statistical performance was measured with accuracy values. The first 100 classification trials were performed by separating 70%-30% of the dataset into training and test data Predictions were made on the test data with the logistic regression model obtained with the training data, and the minimum accuracy value of the model was obtained as 0.62, the maximum accuracy value as 1.00 and the average accuracy value as 0.88. The last 100 classification trials were generated by randomly selecting the degree of separation of the dataset into training and test data between 20% and 80%. Predictions were made on the test data with the obtained model and the minimum accuracy value of the model was 0.48, the maximum accuracy value was 1.00 and the average accuracy value was 0.81.
Keywords: Tree classification, Feature extraction, Logistic regression, Machine learning.