best365网页版登录统计与运筹系

学术报告

Forecast combinations:

modern perspectives and approaches

康雁飞 副教授

（北京航空航天大学）

报告时间： 2023年10月20日 (星期五) 下午3：00-4：00

报告地点 ：沙河主E706

报告摘要：Forecast combinations have flourished remarkably in the forecasting community and, in recent years, have become part of the mainstream of forecasting research and activities. Combination schemes have evolved from simple combination methods without estimation, to sophisticated methods involving time-varying weights, nonlinear combinations, correlations among components, and cross-learning. They include combining point forecasts, and combining probabilistic forecasts. They also include combining multiple forecasts derived from different methods for a given time series, combining the base forecasts of each series in a hierarchy,  and aggregating forecasts computed on different perspectives of the same data. In this talk, I will start from classical forecast combination ideas and present some modern perspectives of combining that can offer substantially improved forecasts on average: 1) combining multiple models: feature-based forecasting, 2) hierarchical forecasting: optimal reconciliation with immutable forecasts, and 3) wisdom of the data: improving forecasting by subsampling seasonal time series. This talk concludes with current research gaps and potential insights for future research on forecast combinations.

报告人简介：

康雁飞，北航经管学院副教授、博导，先后入选北航“卓越百人计划”和北航“青年拔尖人才计划”，现任北航数量经济与商务统计系主任。2014年博士毕业于莫纳什大学，2014-2015年于莫纳什大学从事博士后研究，合作导师为两位澳大利亚科学院院士Kate Smith-Miles 教授和Rob Hyndman教授，2015-2016年就职于百度大数据部。研究方向为时间序列预测及统计计算。共承担科研项目10余项，其中主持国家自然科学基金2项。在European Journal of Operational Research, International Journal of Forecasting等国际权威期刊发表论文28篇。担任中国统计教育学会理事、北京大数据协会理事及超过10个国际学术期刊审稿人。

邀请人： 罗雪