ANN Model for Forest Cover Classification

Abstract

AbstractIn order to develop efficient ecosystem management, an ANN model has been developed to present a factual way to capture futuristic dynamics on forest cover. The forest cover-type Kaggle dataset provided by US Forest Service has been considered for the study. The different activation functions are used to perform varied calculation between the layers of MLP architecture to predict different forest types. The ANN model predicts projection by giving accuracy of 100% with ReLU (ReLU on internal node and Softmax at output node) in comparison with TanH activation function giving 61.21% of accuracy. The results illustrate the toughness and efficiencies of the ANN representation with the combination of ReLU and Softmax. This work offers a consistent means for projecting forest cover and farming yields under provided prospective circumstances, supporting administrative management in consistent land development, management, and protection.KeywordsMultiple layer perceptron (MLP)Rectifier linear unit (ReLU)SoftmaxTanHForest coverArtificial neural network (ANN)Activation function (AF)US Forest Service (USFS)US Geological Survey (USGS)Sustainable forest management (SFM)