摘要
我国天然气进口具有时间短、数据量少的特点,传统的预测方法不能兼顾结果的准确性和实时性.提出了一种基于遗传算法优化极限学习机模型的权重和阈值的新方法,使用随机森林算法评估影响因素的特征重要性,从中选择了影响最显著的6个因素作为模型的自变量;用自变量2006—2018年的历史数据训练经遗传算法优化的极限学习机模型,得到预测精度高的机器学习模型;再使用差分整合自回归移动模型对未来自变量的数值进行预测,将预测结果代入训练好的机器学习模型中,得到未来天然气进口量的预测值.结果显示,我国未来天然气进口量将呈现上升态势,其增长率经历几年下降后将保持平稳.
China’s natural gas import has the characteristics of short time and less data.Traditional forecasting methods can not take into account both accuracy and real-time.In this paper,a new method based on genetic algorithm to optimize weights and thresholds of extreme learning machine predictor is proposed.The historical data of 2006-2018 are selected,the characteristic importance of influencing factors is evaluated by random forest algorithm,and six influencing factors are selected as independent variables of the model.The historical data of independent variables are used to train the model.The GA-ELM model with high prediction accuracy is obtained by training the limit learning machine optimized by genetic algorithm,and then the future independent variables are predicted by using the differential integrated autoregressive moving model,and the predicted results are substituted into the trained GA-ELM model to obtain the predicted value of future natural gas imports.The results show that China’s natural gas imports will show an upward trend in the future,and its growth rate will remain stable after several years of decline.
作者
李宏勋
宫本璞
LI Hongxun;GONG Benpu(School of Economics and Management,China University of Petroleum,Qingdao 266580,Shandong China)
出处
《河南科学》
2020年第4期667-673,共7页
Henan Science
基金
国家社会科学基金项目(12BJY075)
山东省社会科学规划研究项目(16CGLJ46)。
