EVALUATION OF ACCURACY CLUSTERING ZOM BMKG IN NGAWI WITH MULTIVARIATE ADAPTIVE REGRESSION SPLINE (MARS) APPROACHING
Nur Faizah, Bambang Widjanarko Otok, Sutikno Sutikno · 2011
BMKG has made climate classification by making Zon·e Season (ZOM). Based on the results of an evaluation of the climate modeling (rainfall) indicates that there dre several locations (ZOM) had a poor performance of the low accuracy. Some things that allegedly resulted in low accuracy, such as forecasting methods and ZOM determination is not suitable anymore. Therefore, this research aims to evaluate the classification accuracy ZOM BMKG with MARS approach. The process of grouping was using monthly rainfall data in Ngawi on original data (without reduction) and reduction data of factor analysis. Grouping method was using complete linkage and euclidean distance. Furthermore, the results of the grouping are modified through a correction elevation maps, and conducted an evaluation with MARS approach. The results showed that re-grouping ZOM obtained an optimum number of groups as much as 2 groups for the original data and 4 groups for reduction data. Evaluation of the accuracy of the classification ZOM BMKG with MARS approach in the original data and reduction data is 88.9% and 100%. This indicates that MARS is one of the best classification methods in evaluating the classification accuracy ZOM because it has a high classification accuracy.