Volume 9, Number 4

Application of Data Mining Technique to Predict Landslides in Sri Lanka


Karunanayake K.B.A.A.M and Wijayanayake W.M.J.I, University of Kelaniya, Sri Lanka


Landslides are the major natural disaster in hill country of Sri Lanka, which create terrible economical and ecological damages. Therefore, the fast detection is important. Currently in Sri Lanka,predict landslides based on a map reading approach. But a map is limited to specific point in time, and do not take current conditions into account. Therefore, develop a model/tool which has ability to efficiently deal with current situation is important. Hence within this study, prediction models were developed using Decision Tree and Neural Network data mining techniques,based on the data of Badulla and NuwaraEliya districts. Selected Decision Tree model for Badulla district has 96.2963% accuracy level and Nuwara Eliya district has 100% accuracy level. Though Decision tree models were outperformed, Neural Network models also have above 90% accuracy. Therefore, it can be concluded that both data mining techniques are suitableto developlandslide prediction models for Sri Lanka


Landslide, Data mining, Predictive analysis, Plan-Do-Check-Act, Decision tree