基于联合聚类的国家风险集成预测方法研究  

Research on Ensemble Forecasting Approach of Country RiskBased on Joint Clustering Strategy

作　　者：李建平 袁佳鑫 尚书凡 郝俊 LI Jianping;YUAN Jiaxin;SHANG Shufan;HAO Jun(School of Economics and Management,University of Chinese Academy of Sciences,Beijing 100049,China)

机构地区：[1]中国科学院大学经济与管理学院,北京100049

出　　处：《中国石油大学学报（社会科学版）》2023年第5期101-108,共8页Journal of China University of Petroleum (Edition of Social Sciences)

基　　金：国家自然科学基金青年基金(72201265);中国博士后科学基金(2023T160635,2022M723105)。

摘　　要：准确预测国家风险能够有效降低一国在对外贸易中期望收益受损的可能性。然而,国家风险构成复杂、诱因多样且具有突发性,使得精准预测国家风险面临严峻挑战。为此,提出了一种基于BIRCH嵌套AP算法的联合聚类集成预测模型。通过设定移动窗口并使用BIRCH算法压缩并建立聚类特征树,利用AP算法对特征结果进行聚类并选定评估聚类簇;然后,计算单体模型在聚类簇中的预测表现,选取表现最佳的模型作为集成预测的输出。选取沙特阿拉伯、科威特、伊朗、利比亚4个欧佩克成员国的ICRG数据进行仿真实验以验证模型的预测性能,实验结果表明:在不同提前期下所提模型呈现出优异的预测表现,尤其在方向性精度上表现出显著且稳定的优势。Accurate prediction of country risk can effectively reduce the likelihood of a country's expected gains being jeopardized during the foreign trade process.However,due to the complexity of its composition,the diversity of its causal factors and its suddenness,country risk is hard to be accurately predicted.To this end,a joint clustering integrated prediction model based on the BIRCH nested AP algorithm is proposed.By setting a moving window and using the BIRCH algorithm to compress and build a clustered feature tree,the AP algorithm is used to cluster the feature results and select clusters for evaluation;then,the prediction performances of the single models in the clustered clusters are computed,and the best-performing models are selected as the outputs of the integrated prediction.The study selects ICRG data from four OPEC member countries,i.e.,Saudi Arabia,Kuwait,Iran,and Libya,to conduct simulation experiments for validating the forecasting performance of the models.The experimental results show that the proposed model presents excellent bi-directional prediction accuracy under different lead times,especially in directional accuracy which always shows more significant and stable advantages.

关 键 词：国家风险 集成预测 联合聚类 风险预测 模型选择 

分 类 号：F113[经济管理—国际贸易] F224.0[经济管理—世界经济]