Calculating Forest Area Changes Using Different Machine Learning Methods
Bolorchuluun Chogsom1*, Zandanbat Tsog-Urnukh2

1Department of Geography, National University of Mongolia, Mongolia-
2Data specialist, National Statistical Organizaton, Mongolia

Abstract

Forests occupy 9.1% of Mongolia^s total land area, or a very small part. In the era of advanced technology, it is necessary to process remote sensing data with machine learning methods and compare the results. The aim is to find a suitable method for calculating forest changes by processing this forest area change using machine learning methods on active sensing images and comparing it with optical sensing. To calculate forest changes, machine learning methods were compared on active sensing ^Sentinel-1^ satellite images and validated on optical sensing ^Landsat-8^ satellite images. As a result of this research, four machine learning methods used to calculate forest area were tested: support vector machine, K-nearest neighbor method, maximum similarity method, and random forest. Of these, the most similar method was feasible, while the other methods were moderately feasible. This research is innovative in that it combines active sensing with optical sensing and compares the differences between machine learning algorithms using classification methods.

Keywords: Mashine Leaerning, SVM, K-NN, Maximum Likehood, Random Forest

Topic: Topic B: Applications of Remote Sensing