We use the random forest to regress the surface effective temperatures of stars in APOGEE from SDSS DR16 and LAMOST DR6. When the NUV-u, u-g, g-r, r-i, i-J, J-H, H-K, K-WISE_4_5 magnitudes are used as machine learning features, the coefficient of determination of regression are 94.91% in APOGEE and 90.46% in LAMOST. The standard deviation of the prediction and pipeline temperatures are 93.89K in APOGEE and 113.10K in LAMOST. When the NUV-J, J-H, H-K, K-WISE_4_5 magnitudes are used as features, the coefficient of determination of regression are 94.37% in APOGEE and 88.89% in LAMOST. The standard deviation is 96.59K in APOGEE and 119.92K in LAMOST. The J-H magnitudes are the most important feature to predict the effective temperatures, and the NUV-J magnitudes are the second important feature. The NUV-J, J-H, H-K, K-WISE_4_5 magnitudes are from the all-sky survey and can be employed widely to regress the effective temperatures of stars.

我们使用随机森林从SDSS DR16和LAMOST DR6回归APOGEE中恒星的表面有效温度。当将NUV-u，ug，gr，ri，iJ，JH，HK，K-WISE_4_5量级用作机器学习特征时，APOGEE和LAMOST的回归确定系数为94.91％和90.46％。预测和管道温度的标准偏差在APOGEE中为93.89K，在LAMOST中为113.10K。当使用NUV-J，JH，HK，K-WISE_4_5量级作为特征时，APOGEE和LAMOST的回归确定系数为94.37％和88.89％。APOGEE中的标准偏差为96.59K，LAMOST中的标准偏差为119.92K。JH量级是预测有效温度的最重要特征，而NUV-J量级是第二个重要特征。NUV-J，JH，HK，