Medicinal and Aromatic Plants Identification Using Machine Learning Methods

Abstract

In this study, different machine learning (ML) methods were used to classify medicinal and aromatic plants (MAP) namely St. John's wort (Hypericum perforatum L.), Melissa (Melissa officinalis L.), Echinacea (Echinacea purpurea L.), Thyme (Thymus sp.) and Mint (Mentha angustifolia L.) based on leaf shape, gray and fractal features. Naive Bayes Classifier (NBC), Classification and Regression Tree (CART), K-Nearest Neighbor (KNN), and Probabilistic Neural Network (PNN) classification were used as methods. The results indicated that plant species were successfully recognized the average of correct classification rate. The best classification rate on the NBC was taken: training data for classification rate 98.39% and test data classification rate for 98.00% are obtained. ML could be accurate tools for MAP classification tasks.

An Accurate HOG based Exemplar Pyramid Method for Image Classification of Astragalus L. Taxa

“…Also, Sun et al (2017) used deep learning models to design plant classification in the natural habitats and they found that the achievement ratio of model was determined 91.78% (Sun et al, 2017). Similarly, Kayhan and Ergün (2020) classified the medicinal and aromatic plants by using several machine learning techniques and their results showed that plants were diagnosed in the correct classifications.…”

Comparison of the performance of different learning algorithms in leaf feature extraction and recognition using convolution neural network

“…This database has also been classified with traditional methods (NBC, CART, KNN, and PNN methods), and it has been observed that a long processing time is required. 12 Despite this drawback, the CNN deep learning method provides high-accuracy test performances with three different training algorithms. In all three methods, none of these leaves is confused with the "2-thyme" leaf.…”

“…Traditional methods 12 Manual Manual Complex Proposed method Automatic Automatic Basic combines algorithms from a recurrent neural network and CNN and classifies the deep features extracted with the CNN model using classical classifiers. Our preliminary study 12 used NBC, CART, KNN, and PNN methods to classify the leaves from the database of medicinal and aromatic plants using their shape, and gray and fractal features. Liu et al 13 classified the Flavia leaf database, consisting of 32 species and 4800 leaves, with a 10-layer CNN.…”

“…These are the CNN model and four different transfer learning models, which are based on fine tuning and deep feature extraction, a cross dataset fine tuning model and a model that combines algorithms from a recurrent neural network and CNN and classifies the deep features extracted with the CNN model using classical classifiers. Our preliminary study 12 used NBC, CART, KNN, and PNN methods to classify the leaves from the database of medicinal and aromatic plants using their shape, and gray and fractal features. Liu et al 13 classified the Flavia leaf database, consisting of 32 species and 4800 leaves, with a 10‐layer CNN.…”

Classification of Aviation Accidents Using Data Mining Algorithms

“…During calculation of the values of said concepts, the comparison of the estimated and available data is taken into account [33] Using the confusion matrix given in Figure 2, the accuracy values of the classification algorithms can be calculated. The precision statement is the ratio of the number of correct and positive estimated samples as class 1 to the number of estimated samples as class 1, as indicated in Equation ( 7) [34]. Sensitivity is defined as the ratio of the number of positive samples correctly classified in Equation ( 8) to the total number of positive samples.…”