Perencanaan dan Perancangan Pembangunan Prototipe Evaluasi dan Feedback

Tan P, Steinbach M, Kumar V .2006. Introduction to Data Mining. Addison- Wesley. Liu P, Zhou D, Wu N. 2007. Varied Density Based Spatial Clustering of Application with Noise. Proceedings of IEEE Conference ICSSSM. LAMPIRAN Lampiran 1 Gambar Keseluruhan penggerombolan DBSCAN warna mengilustrasikan perbedaan penggerombolan 59 Lampiran 2 Gambar Keseluruhan penggerombolan DBSCAN dengan data perbulan warna mengilustrasikan perbedaan penggerombolan 61 ABSTRACT UTSRI YUSTINA PURWANTO. Spatial Hotspots Clustering of Forest and Land Fires using DBSCAN and ST-DBSCAN. Under direction of BABA BARUS, and HARI AGUNG ADRIANTO. Forest and land fire has become international important environmental and economic issue for the last several years. For Indonesia and some neighboring countries, it also produces huge amount of smog and air pollution causing economic, environmental and health problems. The objective of this research is to find hotspot clustering pattern using DBSCAN dan ST-DBSCAN. There is a possibility that the hotspot is not spread randomly but naturally gather in some area to form clusters based on proximity of distance and time. DBSCAN and ST- DBSCAN are density based clustering in spatial data mining. The advantage of these methods is the clusters form can be more flexible, especially when applied on a large data size. However, DBSCAN cannot differentiate two adjacent clusters with different density while ST-DBSCAN can. As the results, DBSCAN clustering detected 38 hotspot clusters with 6 noises while ST-DBSCAN detected more. Since ST-DBSCAN is an extension of DBSCAN algorithm, ST-DBSCAN can process both of spatial and non-spatial data by using Eps 1 for spatials attributes and Eps 2 for non-spatial attributes, such as time. With this algoritm, there are 147 hotspot clusters and 149 noises. The biggest cluster are located in Musi Rawas, Muara Enim, Musi Banyuasin, dan Ogan Komering Ilir. Furthermore, this research is expected to identify hotspot clustering patterns and behaviors and to produce useful information to evaluate and mitigate forest and land fires hazards and also generate it’s prototype. Keywords : Spatial Data mining, Forest and Land Fires, Clustering, DBSCAN, ST-DBSCAN