IMPLEMENTATION OF MULTI-CLASS GRADIENT BOOSTING TO CLASSIFY ANIMAL SPECIES IN ZOOS

Abstract

Animals are one of the living things that have various types. Grouping types of animals based on similarities and differences in characteristics owned is one of the important activities carried out To make it easier to compare, recognize, study certain types of animals and be able to find out kinship relationships between animals, So if a new type of animal is found that does not yet have a name, it will be easier for us to give a name to the animal based on the type and based on the group. In research on the classification of animal species in zoos that have multi-class, the best classification is obtained by applying gradient boosting parameters with n_estimators of 50, max_depth 3, sub-sample of 1.0, learning rate of 0.1, and using criterion friedman Mse. And by implementing Split validation or division between training data by 80% for training data and 20% for testing data. The results stated that the proposed model was better than some other models that had also been tested with an accuracy value of 93.75%, recal of 94%, precision of 96% and MSE to measure the average magnitude of error in a series of classifications of 12.5%, the smaller the MSE value, the better it would be in classifying.