ADAPTIVE SYNTHETIC IMPLEMENTATION ON RANDOM FOREST IN ARCHIPELAGIC FISHING PORT OF PEMANGKAT

Abstract

Random Forest is one of the classification methods employed in data mining. One of the problems in data mining classification is the problem of unbalanced class data This phenomenon arises when the data classes utilized do not have identical instances. Imbalance class data causes the classification results to be biased towards the majority class. Adaptive Synthetic (ADASYN) can be used to deal with this problem. ADASYN generates synthetic data by assigning different importance of minority class samples and then producing synthetic data with similar characteristics. The implementation of ADASYN is suitable for fishery production data, which will experience the problem of unbalanced class data. Fish production is part of the measured fishery. This study aims to classify the value of measured fishery production at PPN Pemangkat through Random Forest Classification using ADASYN to handle the imbalance class data problem and compare the results with those without ADASYN implementation. This study uses four predictor variables which include fishing gear types (), number of trip days (), number of crew (), and the total weight of fish () with production value as response variable (). Accuracy, precision, recall, specificity, and G-mean are the model performance indicators used. The results showed that ADASYN successfully handles the problem of unbalanced class data in Random Forest classification. Accuracy is increased from  to , Specificity is increased from  to , Precision from  to , and G-Mean from  to . The  decrease in recall is negligible due to the small amount, so the Random Forest classification with ADASYN is better than without ADASYN