Multiclass classification of 7 different forest cover types found in the Roosevelt National Forest using Support Vector Machines (SVMs) and Multilayer Perceptron's (MLPs).
This study aims to compare and contrast the performance of Support Vector Machines (SVM) with Multilayer Perceptron (MLP) models at predicting different forest cover types. The prediction is a multiclass classification problem, based on environmental data representing seven imbalanced forest cover types found in the Roosevelt National Forest.
Forest cover type prediction is important for land management agencies as it reflects the ecological balance and diversity of an area. Alternatives to computer-based modelling involve direct recording and estimates based on remote sensing data – however both approaches have limitations in terms of cost, time and prediction accuracy. It may also be necessary to predict cover distribution in areas that are not legally accessible, hence the need for accurate models.
For more details please see the pdf report above.
Data source: https://archive.ics.uci.edu/ml/datasets/covertype