Using Various Machine Learning Models
- To view the Code with Output check this Link - https://ayonija.github.io/ForestCoverDataStudy/Forest_CoverType
Abstract— The study evaluated four wilderness areas in the Roosevelt National Forest, located in the Front Range of northern Colorado. We use a series of mining concepts including both linear and non-linear algorithms to predicted forest cover type. Compared with the linear models, non-liner models satisfy more effectively the classification. This report deals with the study of how these models respond to forest cover type problem. We are using RStudio to visualize these models. The accuracy rate on test data have been studied and compared with each model. Also got closer to the performance by the updated models used to classify forest categories.
INTRODUCTION— We attempt to predict the predominant type of tree in sections of wooded area. Understanding forest composition is a valuable aspect of managing the health and vitality of our wilderness areas. The study area includes four wilderness areas located in the Roosevelt National Forest of northern Colorado USA. Each observation is a 30m x 30m patch. This project focuses to predict an integer classification for the forest cover type. The seven types are: 1 - Spruce/Fir 2 - Lodgepole Pine 3 - Ponderosa Pine 4 - Cottonwood/Willow 5 – Aspen 6 - Douglas-fir 7 – Krummholz. These areas have experienced relatively little direct human management disturbances, thus the current composition of cover type within them are primarily a result of natural ecological processes, rather than the product of active forest management.
Reference and DataSource - https://www.kaggle.com/c/forest-cover-type-prediction/data