Automatic identification of five Vatica species (Vatica spp.) based on some leaf morphological characters and machine learning algorithms
DOI:
https://doi.org/10.55981/bkr.2023.740Keywords:
automatic identification, Dipterocarpaceae, leaf character, machine learning, VaticaAbstract
Resak (Vatica spp.) is a genus of the timber family, Dipterocarpaceae, with several species categorized as threatened. One of the most important components of conservation efforts is accurate species identification. This study aimed to identify the morphological traits of resak leaves, the similarity between species, and the effectiveness of five machine learning methods in automatically recognizing resak species. The measured morphological characters included the color, size, shape, and texture of the leaves of five species of Vatica growing at Bogor Botanic Garden. The differences in the mean value of each morphological character were analyzed using analysis of variance and the Tukey test. Morphological diversity and similarity were analyzed using Principal Component Analysis and Cluster Analysis. Automatic identification was conducted using five machine learning algorithms, namely BayesNet, K-Nearest Neighbor, Artificial Neural Network, Random Forest, and Support Vector Machine. The results showed that leaf morphological characters (color, size, shape, and texture) in five species of resak have significant differences. All morphological characters significantly affect the differences in resak leaf characteristics. At an 80% similarity level, the five species were grouped into three clusters, namely cluster I (V. granulata, V. pauciflora, V. venulosa), cluster II (V. bantamensis) and cluster III (V. rassak). The best machine learning algorithm for identifying resak species based on leaf morphological characters is K-Nearest Neighbor with an overall accuracy value of 0.92, a Kappa coefficient of 0.90, an average precision of 0.93, and an average recall of 0.92.
