Classification of forest cover-type using ensemble of decision tree, random forest and K nearest neighbor

Natural resources are an essential part of any eco-system. Forest, a vital natural resource highly affects agriculture and other bioclimatic events. Forest cover type classification using Machine learning-based classifiers has been an active area of study in the research community. The article examines the performance of the ensemble method in classifying forest cover types. The ensemble method combines predictions from Decision Trees, Random Forest, and K Nearest Neighbor algorithms to generate the final prediction. The forest cover type (FC) dataset is taken from the UCI machine learning repository. The empirical results show that the proposed ensemble technique helps achieve an accuracy of 97.10% on 10-fold cross-validation, which is significantly better than the research works available in the literature. The obtained results also show much improvement in terms of other performance metrics like F1 score, Precision and Recall, thereby signifying the potential of the proposed work.