<![CDATA[Remote sensing technologies have been used for monitoring basal stem rot (BSR) diseases in oil palm plantations. The classification performance could be improved with machine learning algorithms. The success key of BSR treatment in culture technic treatments to prolong oil palm productivity ages is early detection of BSR disease . The selection of classification models of machine learning algorithms will be affecting accuracy and classification processes time. Random forest (RF) of model classification based on several previous research has a high performance of accuracy for BSR disease classification in the oil palm plantation. Today, there were fifteen models of the variance of the random forest algorithm. The research goals were to compare the accuracy and time consumed for the fitting model from 15 model’s classification of RF group in case of identity and classify the BSR disease in the oil palm plantations. Descriptive analysis was adopted to explain the difference accuracy and time for the fitting model of fifteen model classification. The results showed the highest accuracy and Kappa value (0,914 dan 0,815) was the oblique random forest (ORF) with partial least squares (PLS) method and random ferns (Rferns) was the method with the lowest accuracy and Kappa value (0,657 dan 0,334). The highest time consumed for the fitting model was RF rule-based model (rlb) with 12.993,23 seconds and the shortest time was parallel random forest (PRF) with 13,54 seconds. The accuracy and Kappa values and time consumed for the fitting model as a consideration to selecting model classification of healthy and infected BSR in oil palm plantations.]]>
Copyrights © 2020
