Land Cover Classification Using Linear Support Vector Machines

Abstract

GIS has been an effective tool in identifying and recognizing urban patterns. Various techniques like Support Vector Machines, artificial neural networks have been used with GIS to classify the patterns for urban analysis. Liblinear has emerged as another effective tool which produces results in much lesser time without compromising the accuracy. In this thesis the datasets used were extracted using GIS. The datasets were from the Ohio state counties namely the Delaware, Holmes, Mahoning and Medina counties. Each had over a million records and contained seven independent variables related to urban development and a class label which denotes the urban areas versus the rural areas. Using Liblinear, Libsvm, Rapid Miner and Weka some experiments were carried out over smaller datasets and the results have been shown. It can be seen that Liblinear is as effective as Libsvm while the latter takes much longer time for producing the results. The results can help identify geographical patterns related to urban land use.