Prediction of Medicinal Leaves Using Machine Learning Algorithms

Medicinal plants are used for centuries by various cultures worldwide to treat a wide range of ailments and conditions. As there are vast numbers of medicinal plant species it will be challenging for a layman to identify them accurately. In such a situation, the experts' knowledge is sought many times. Unfortunately, lack of expertise, commercial motive and reduced availability of medicinal plants sometimes result in maltreatment due to improper plant identification. The advent of machine learning techniques and their uniqueness motivated the proposal of models that help to classify and predict medicinal plants based on their leaf images. A dataset of 30 different types of medicinal plant leaves sourced from Kaggle is used for this research. In this research a total of 19 features including the texture and shape feature of the leaves are considered extracted through image processing. Six machine learning models such as Extreme Gradient Boosting, Decision Tree, Support Vector Machine, Random Forest, K-Nearest Neighbour, and Naïve Bayes and three ensemble models namely XGRF, XGDT and XGNB are proposed and developed using Python. XGRF achieves an AUC of 93.21% for train and test splits 80%:20% and cross-validation is employed to ensure the robustness of the model. Of the 19 features, through feature importance 8 features are observed to dominate in leaf identification. With these 8 features, XGRF is found to achieve an AUC of 93.19%. The result indicates the feasibility of the model to automatically classify medicinal leaves.