一种基于特征选择与迁移学习的数据预测方法

陈通宝,温亮明,黎建辉

A Data Prediction Method Based on Feature Selection and Transfer Learning

Tongbao Chen,Liangming Wen,Jianhui Li

表3 三种方法分别在8个国家的测试集中得到的MAE和RMSE

Table 3 MAE and RMSE obtained by three methods in the test set of eight countries

目标国家 方法 MAE RMSE Bahrain 均值插补 17.910465 17.910814 SVR 0.243053 0.296189 TLM-SVR(Finland-Bahrain) 0.23997 0.288697 Botswana 均值插补 29.374419 29.456773 SVR 5.976785 6.945181 TLM-SVR(Australia-Botswana) 0.652398 0.882015 Brazil 均值插补 49.648837 49.655030 SVR 2.526161 2.575070 TLM-SVR(Costa Rica-Brazil) 0.813297 0.864135 Colombia 均值插补 26.443023 26.448932 SVR 1.781718 1.782010 TLM-SVR(Canada-Colombia) 0.75966 0.760152 Costa Rica 均值插补 14.148256 14.153092 SVR 2.455805 2.486353 TLM-SVR(Poland-Costa Rica) 0.407760 0.488031 Romania 均值插补 19.705233 19.708832 SVR 1.922851 1.959395 TLM-SVR(Japan-Romania) 0.330607 0.386404 Uruguay 均值插补 18.742442 18.743426 SVR 1.888115 1.888998 TLM-SVR(Chile-Uruguay) 0.102324 0.131846 El Salvador 均值插补 4.059302 4.060842 SVR 2.158907 2.159165 TLM-SVR(Austria-El Salvador) 0.105168 0.116916