FOREST COVERTYPE CLASSIFICATION
MATEJ KVAŠŠAY
PROJECT DESCRIPTION
PROJECT GOALS
▸ Train various models for classifying forest cover into 7 classes
▸ Evaluate performance, compare models and investigate which attributes contribute the most
▸ Demonstrate usage of methods introduced in this course
DATASET
▸ “Covertype data set” from UCI ML repository (1998)
▸ Over 580000 data instances
▸ 54 cartographic attributes
▸ Multivariate data - quantitative (elevation, slope, aspect, hillshade) and binary (soil types)
PROPOSED APPROACH
▸ Statistical analysis of attribute values
▸ Feature selection
▸ Training SciKit-Learn models with multiple configurations
▸ Cross validation