COMPARISON OF DOUBLE RANDOM FOREST AND LONG SHORT-TERM MEMORY METHODS FOR ANALYZING ECONOMIC INDICATOR DATA

Abstract

The performance of machine learning in analyzing time series data is being widely discussed. A new ensemble method Double Random Forest (DRF), which considers supervised learning currently developed. This method has been claimed to be able to improve the performance of Random Forest (RF) if the data is under-fitting. Another machine learning method, Long Short-Term Memory Networks (LSTMs) have capability to analyze nonlinear data. Since the study compare both methods has not been existed in literature, it is interesting to compare the performance of both methods using Indonesian data, especially economic indicator data which have been found to be under-fitting, non-underfitting, and nonlinear data. The indicators used in this study are Export, Import, Official Reserves Asset, and Exchange Rate data. The results showed that overall, the LSTMs method outperforms DRF method in analyzing the data.