自回归积分移动平均
计算机科学
组分(热力学)
政府(语言学)
生产力
平均绝对百分比误差
时间序列
人工智能
均方误差
机器学习
计量经济学
数据挖掘
人工神经网络
统计
经济
宏观经济学
数学
物理
热力学
语言学
哲学
作者
Emmanuel Dave,Albert Leonardo,Marethia Jeanice,Novita Hanafiah
标识
DOI:10.1016/j.procs.2021.01.031
摘要
Export is an important factor that keeps the economy of a country going. Local export forecast guides government for a better policy making, local productivity measurement and international trade preparation. This research aims to provide governments with an accurate prediction of Indonesia’s future exports by building an integrated machine learning model. This hybrid learning model is compared with individual learning models to obtain the most accurate model. The hybrid model integrates ARIMA and LSTM models based on their specialties, where LSTM was applied on the non-linear component of the data and ARIMA was applied on the linear component of the data. The hybrid (LSTM-ARIMA) model achieves the lowest error metrics among all the tested models. It succeeds to outperform the other standalones models, achieving a MAPE value of 7.38% and a RMSE of 1.66 × 1013. Lastly, the entire dataset is used to train the final hybrid model to forecast Indonesia’s exports one year ahead. This forecast can be used by government in guiding them in decision making to foster the future economy.
科研通智能强力驱动
Strongly Powered by AbleSci AI