CLASSIFICATION MODEL USING ANFIS ALGORITHM FOR ESTIMATION OF LANDCOVER CLASS AMBIGUITIES (Study Case: Mixed Garden and Private Forest)
octo sani, Muhamad Buce Saleh, Ade Komara Mulayana · Jurnal Pengelolaan Sumberdaya Alam dan Lingkungan (Journal of Natural Resources and Environmental Management) · 2019
The spatial information is importance in natural resources management for sustainable development. It is a land cover map which is update and comprehensive, to describe phenomenon of land use changed in Indonesia. This research was done by applying the classification of land cover ambiguities between mixed garden and private forest using ANFIS algorithm. This algorithm compute spectral, physic and socio-economy factors in to fuzzy model. The best classification model which chosen is a model scenario with K-fold 3 as training data by 80 data rows, with type of membership function on triangle distribution, with the number of membership functions as follow; the distance to settlement is at 4 mf (near, medium, far, very far), the distance to rice field is at 4 mf (near, medium, far, very far), the elevation is at 2 mf (low and high), the near infrared band is at 2 mf (low and high), the thermal infrared band is at 2 mf (low and high). It shows that error value is downward trend and stable at 0.311646 and learning iteration at 3108. It give the overall accuracy of data test at 85%, and data validation at 69.23%. This model is adequate to data input characteristic, so the degree of membership calculated firing strength that was satisfied to generate the best of premise statement on rule classifier.