Machine learning in the field of agriculture is expanding as an emerging technology. The main objective for any machine learning algorithm in case of classification is to select relevant features for reducing the computational load. However, uncertainty being present in features due to insufficient measurement or imprecise information captured from various experts invites new techniques to select the relevant features. Neutrosophic set theory is approached to select the features in presence of uncertainty. The present work emphasizes the role of Neutrosophic set theory in machine learning as feature selection tool for classifying paddy seed. The proposed work is done with two phases. In the first phase, extraction of feature is carried out using traditional image processing technique and subsequently important features are selected using Neutrosophic set theory. Single valued Neutrosophic concept is used to score of the Neutrosophic number assigned to features. In the second phase, classification algorithm is applied on the reduced dataset to check the performances of the selected features in terms of classification parameter values. To perform the experiment of the proposed method, dataset is collected from the Kaggle and UCI repository. Results are in good agreement with physical reality implying that the proposed model is efficient one for seed classification.