Image Classification of Tourist Attractions with K-Nearest Neighbor, Logistic Regression, Random Forest, and Support Vector Machine

Abstract

K-Nearest Neighbor (KNN), Logistic regression (LR), Random Forest (RF), and Support Vector Machine (SVM) are four methods of identification. The methods are widely used in various research in data mining, especially classifications in recent years. We have used the four classification methods in the study to classify images of five natural attractions, namely Danau Toba (North Sumatra), Nusa Penida (Bali), Raja Ampat (West Papua), Tanah Lot (Bali), and Wakatobi (Southeast Sulawesi). Our research results have concluded that the Logistic Regression method's performance has the best performance in classifying natural images as done in this research. The LR method can classify images that other methods such as kNN, SVM, and RF cannot be correctly classified. However, SVM also shows good performance by only making one error in the classification results; it can even be corrected using the Linear Kernel. In general, it is shown that the LR method has the highest precision value of 100%, followed by the method of kNN and SVM with a precision of 91.9% and RF with a precision of 81.9%. Variations of the variables used in the experiment also determine each method's precision. Chebyshev Metric has the highest precision value in the kNN method, and Ridge Regularization has the highest precision value in the LR method. The number of best on the RF method is 11, and Linear Kernel is the Kernel that gets the best precision value on the SVM method.