Decision tree classification model for land use and land cover mapping- a case study

Decision Tree classification algorithms have significant potential for land cover and land use mapping, few studies have already assessed the use of Decision Tree Classifiers. Decision Tree Classifiers have several advantages for remote sensing applications by virtue of their relatively simple, explicit, and intuitive classification structure. In this paper, the Standard Decision Tree algorithms used for Land Cover and Land Use mapping was evaluated and compared using satellite data. Here multi-spectral IRS-1D/LISS III image of Heggadadevanakote taluk was used as an experimental data. Classification rules were derived from the spectral image using J48, BFTree, REPTree and Simple Cart algorithms with the same set of training samples. Classification done by using these rules is known as knowledge based classification and the results of these classifiers were compared and evaluated based on True Positive, False Positive, Prediction Accuracy and Learning Time metrics. Among these, J48 performed the best in all aspects and had a Prediction accuracy of 97.34% and Kappa statistics of 0.9685. Further the Decision Tree classified image generated using the J48 algorithm based rules produced an overall accuracy of 87.11% and kappa of 0.8515.