NEAREST NEIGHBOR ANALYSIS OF LAND COVER CHANGES BASED ON GEOGRAPHICALLY WEIGHTED-BINARY LOGISTIC REGRESSION MODEL (CASE STUDY: NORTH WEST JAVA DEVELOPMENT REGION)

• Main Author: Rahma Muzdalifah, Qonita
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/40174
• Institution: Institut Teknologi Bandung

Land is one of the important resources that can be used to supply the needs of human life. Uncontrolled land utilization will cause land cover change phenomenon. Land cover change phenomenon can be analyzed by using a model. To get an accurate result, the selection of models in the analysis of land cover change must be based on the characteristics of land cover change phenomenon itself. Land cover change is a binary phenomenon and strongly related to the local characteristics of region. One of the right models that can be used according to those characteristics of land cover change is Geographically Weighted - Binary Logistic Regression (GW-BLR) model. The important thing that needs to be considered in the use of GW-BLR model is GW-BLR model only uses a set of nearest neighbor data with certain bandwidth to estimate the model. The bandwidth value of nearest neighbor data in the GW-BLR model will determine the level accuracy of the produced-model. Therefore, it is necessary to identify the optimal value of nearest neighbor bandwidth for each point of land cover change data in the North West Java Development Region based on GW-BLR model. The results of analysis denote each point of land cover change data in the North West Java Development Region has a different value of optimal nearest neighbor bandwidth. It indicates that the land cover change phenomenon in North West Java Development Region has different local variation. The land cover change modeling results in the North West Java Development Region using GW-BLR method based on the optimal value of nearest neighbor bandwidth have the overall accuracy value of 75.44%.