K-Means clustering visualization on agriculture potential data for villages in bogor using mapserver

Abstract

Abstract. Central Bureau of Statistics have conducted same surveys for vil/age potential which convey data ofpotential or condition including social condition, economical condition, land utilizing. structure and infrastructure up to level of village. This research aims to analyze 5808 records of the vil/age potentiat data especially Ihose which relaled la agriculture. 11could be done by applying data mining techniques in order la gel information or knowledge as a decision support for agriculture sector developmenl. This research uses one of data mining techniques that is clustering using KMeons algorithm and result of clustering is visualized in a form of web based geographical information sys lem. The data used in this research is data of villages especially related to rice jield and other attributes /01' non-agricultural in West Java in 2006 especial/y in Bogor. The results of this research are mean values from each cluster and visualization Jar each componenI of the cluster in a form of web based geographical information system. The best clustering is reached when size of cluster is 4 and random seed is 20. having lalai value of SSE (sum of square error) 1.6702 with a distribution that is nol divided well in each cluster. Members of cluster 0 and cluster 3 are regions with rice field that quite narrow (mean value of attribute area 0/ rice field is 56.3103 Ha in cluster 0 and is 274.7950 Ha in cluster 3). Members ofcluster 1 are regions having large area ofrice field (mean value is 2756 Ha). Members ofcluster 2 are large nonagricultural regions (mean value is 8650.2 Ha).