The following information was submitted:
Transactions: WSEAS TRANSACTIONS ON COMPUTERS
Transactions ID Number: 52-201
Full Name: Retno Kusumaningrum
Position: Lecturer
Age: ON
Sex: Female
Address: Jl. Prof. H. Soedarto SH Tembalang, Semarang,Central Java
Country: INDONESIA
Tel:
Tel prefix:
Fax:
E-mail address: retno_ilkom@undip.ac.id
Other E-mails: retno_309352@yahoo.co.id
Title of the Paper: Land Use Pattern Identification in Indonesian Territory Based on Image Mining Technique
Authors as they appear in the Paper: Retno Kusumaningrum, Katmoko A. Sambodo, Wiweka Hartoyo, Dina Chahyati, Aniati Murni
Email addresses of all the authors: retno_ilkom@undip.ac.id, katmoko_ari@yahoo.com, wiweka@yahoo.com, dina@cs.ui.ac.id, aniati@cs.ui.ac.id
Number of paper pages: 10
Abstract: This paper presents the identification of land use pattern in Indonesian territory and evaluates two types of decision tree classifiers, J4.8 algorithm and NBTree classifier for obtaining the feature template of each pattern. The construction of this classification model was implemented using open source software, WEKA (Waikato Environment of Knowledge Analysis). The evaluation of the resulting classification model is done based on the value of the accuracy, sensitivity and error rate. There are seven patterns of land use that have been identified, namely corridor (is associated with the river), block (shows the fish farms), cubicles (represents the plantation area at the time of planting), network (is related to built area in cities), geometrics (is related to built area in big cities), holes (is associated with dam/lake), and diffuse (represents paddy field at the time of planting). The comparison of using J4.8 and NBTree shows that the NBTree does not always h!
ave better performance than J4.8. In fact, some decision trees that are built using J4.8 have a better performance than the decision tree built using NBTree. The feature template for each land use pattern was obtained by a decision tree that was built using J4.8 algorithm. The combination of landscape metrics that are measured using the area-weighted mean and the color composition (average red, green, and blue) are forming a dataset. The accuracy value of recognizing a pattern is 94.81%, with the sensitivity value of 81.82% and the error-rate value of 5.19%.
Keywords: Land use pattern, Decision trees, J4.8 algorithm, NBTree classifier, Landscape metric, 10-fold cross validation, Accuracy, Sensitivity, Error-rate
EXTENSION of the file: .doc
Special (Invited) Session:
Organizer of the Session:
How Did you learn about congress:
IP ADDRESS: 114.124.10.199
