Developing Land Use Classification Models Using Machine Learning Techniques: The Case of Asella Town

Abstract

Land is the most vital resource on earth plays significant role through Economic, Social, Political, Cultural and technical framework dimensions within must good land management and administration. By the 818/2014, urban land holding adjudication and registration need land reform. The reforms include urban land use changing and transformation up to the social interest of the administrative area. The Asella town 31% proposed structural plan contradict with the existing situation. In such case reclassification, proposed structural plan is necessary according to governmental urban plan policy. The roe of land to Asella town to bring sustainable development of the town is crucial. The study preformed balancing using SMOTE over sampling and implemented stratified kfold for cross validation score. Then the model designed using parametric KNN, RF, DT and SVM and none parametric NB machine learning algorithms used to evaluate the reclassification performance of the designed model. According the result observed from the study KNN, RF and DT score high efficiency more than 99% of accuracy, precision, recall and f1-score. SVM and NB scored lower efficiency compared to other models. In general, the study's findings suggest that machine learning prediction models can identify land use classes and according to the experiment, KNN, RF and DT models are most suitable for land use categorization in respect of Asella town land-use administration undertakings.