基于电力大数据的中国工业增加值现时预测

doi:10.11959/j.issn.1000-0801.2021143

Abstract

Abstract:

Industrial added value is an important indicator to measure the operation of the real economy.In order to fully mine the value of power data in the current macroeconomic nowcasting, so as to serve the government policy making, the Bagging and Boosting algorithms in machine learning were applied to nowcast industrial added value based on electric power data as well as traditional statistical data.Firstly, traditional statistical data can significantly improve the forecasting effect of the ARIMA model.Secondly, the nowcasting ability of electric power data depends on the selection of electric power indicators, and the proper electric power index is helpful to predict industrial added value more timely and accurately.Thirdly, the prediction ability of electric power data to the industrial added value in the current period may be lower than that in the future, which means the power data is more likely to be used to predict ahead of time.

Key words: industrial added value, nowcasting, electric power data

CLC Number:

F224

Fang PENG, Gaoqun PENG, Yaru QI, Tiantian LIU, Xiaolei ZHOU. Nowcasting of China’s industrial added value based on electric power big data[J]. Telecommunications Science, 2021, 37(7): 115-125.

Figures/Tables 10

References 11

[1]	尹德才, 张文 . 当前宏观经济现时预测研究综述[J]. 统计与信息论坛, 2020,35(1): 121-128.
	YIN D C , ZHANG W . A review of current macroeconomic nowcasting research[J]. Statistics ＆ Information Forum, 2020,35(1): 121-128.
[2]	顾光同, 许冰 . 中国工业增加值的半月预报:基于宏观月度数据[J]. 系统工程理论与实践, 2018,38(8): 1983-1993.
	GU G T , XU B . Half-month forecast of China’s industrial added value: based on the macro monthly data[J]. Systems Engineer-ing-Theory ＆ Practice, 2018,38(8): 1983-1993.
[3]	王雷, 聂常虹 . 中国债券利差对宏观经济指标的预测能力研究[J]. 管理评论, 2019,31(5): 66-76.
	WANG L , NIE C H . Research on how predictive Chinese bond spreads is of macroeconomic indices[J]. Management Review, 2019,31(5): 66-76.
[4]	STOCK J H , WATSON M W . Forecasting using principal components from a large number of predictors[J]. Journal of the American Statistical Association, 2002,97(460): 1167-1179.
[5]	STOCK H , WATSON W . Generalized shrinkage methods for forecasting using many predictors[J]. Journal of Business and Economic Statistics, 2012,30(4): 481-493.
[6]	BAI J , NG S . Boosting diffusion indices[J]. Journal of Applied Econometrics, 2009,24(4): 607-629.
[7]	BAI J , NG S . Determining the number of factors in approximate factor models[J]. Econometrica, 2002,70(1): 191-221.
[8]	BREIMAN L . Bagging predictors[J]. Machine Learning, 1996,24(2): 123-140.
[9]	INOUE A , KILIAN L . How useful is bagging in forecasting economic time series? A case study of U.S.consumer price inflation[J]. Journal of the American Statistical Association, 2008,103(482): 511-522.
[10]	KIM H H , SWANSON N R . Forecasting financial and macroeconomic variables using data reduction methods:new empirical evidence[J]. Journal of Econometrics, 2014(178): 352-367.
[11]	BüHLMANN P , YU B . Boosting with the L₂loss:regression and classification[J]. Journal of the American Statistical Association, 2003(98): 324-339.

Metrics

Recommended 0

No Suggested Reading articles found!

	方法				移动窗口					累计窗口
	方法		模型1	模型2	模型3	模型4	模型5	模型1	模型2	模型3	模型4	模型5
有滞后		DI	0.596 0	2.715 7	1.632 2	1.898 0	0.594 4	0.590 3	2.654 6	1.525 1	1.947 9	0.579 1
		Bagging	0.605 4	2.569 4	1.677 8	2.077 2	0.621 1	0.593 5	2.546 2	1.541 1	2.016 5	0.577 5
		Boosting	0.912 0	3.162 2	1.202 9	1.319 3	0.899 5	0.928 5	2.223 9	1.089 6	1.389 5	0.918 9
无滞后		DI	0.596 0	1.563 0	1.448 2	1.600 8	0.599 1	0.590 3	1.537 9	1.435 8	1.596 9	0.595 1
		Bagging	0.605 4	1.537 9	1.462 4	1.627 5	0.611 7	0.593 5	1.509 2	1.435 8	1.585 8	0.595 1
		Boosting	0.912 0	1.624 4	1.369 6	1.278 4	0.908 9	0.928 5	1.761 3	1.402 3	1.442 2	0.926 9

	方法				移动窗口					累计窗口
	方法		模型1	模型2	模型3	模型4	模型5	模型1	模型2	模型3	模型4	模型5
有滞后		DI	0.596 0	1.158 9	0.748 5	1.390 1	1.118 0	0.590 3	2.030 9	0.574 3	1.472 5	0.947 6
		Bagging	0.605 4	1.135 3	0.773 7	1.369 6	1.116 5	0.593 5	1.984 6	0.582 3	1.437 4	0.950 8
		Boosting	0.912 0	1.366 5	0.943 5	1.122 7	1.280 0	0.928 5	1.900 0	0.847 1	1.439 0	1.233 2
无滞后		DI	0.596 0	1.622 8	0.891 6	1.577 2	0.596 0	0.590 3	1.268 3	0.828 0	0.893 4	0.591 9
		Bagging	0.605 4	1.517 4	0.860 1	1.382 2	0.613 3	0.593 5	1.233 2	0.749 8	0.740 2	0.593 5
		Boosting	0.912 0	1.226 5	0.838 1	1.003 2	0.910 5	0.928 5	1.515 6	0.855 1	0.457 9	0.930 1

	方法		移动窗口					累计窗口
	方法		模型1	模型2	模型3	模型4	模型5	模型1	模型2	模型3	模型4	模型5
有滞后		DI	0.596 0	2.133 8	0.418 3	1.520 6	0.589 7	0.590 3	1.577 8	0.408 4	1.660 7	0.579 1
		Bagging	0.605 4	2.102 4	0.405 7	1.165 2	0.605 4	0.593 5	1.550 7	0.408 4	1.322 5	0.574 3
		Boosting	0.912 0	1.108 6	0.710 8	0.897 9	0.894 7	0.928 5	1.025 8	0.743 4	1.054 5	0.950 8
无滞后		DI	0.596 0	2.022 2	1.426 2	1.718 7	0.600 7	0.590 3	1.994 2	1.344 9	1.877 7	0.593 5
		Bagging	0.605 4	1.967 2	1.442 0	1.718 7	0.611 7	0.593 5	1.938 3	1.344 9	1.802 7	0.595 1
		Boosting	0.912 0	1.953 0	1.341 3	1.191 9	0.907 3	0.928 5	2.136 2	1.348 1	1.435 8	0.925 3

	方法		移动窗口					累计窗口
	方法		模型1	模型2	模型3	模型4	模型5	模型1	模型2	模型3	模型4	模型5
有滞后		DI	0.596 0	2.146 4	0.407 3	1.567 8	0.589 7	0.590 3	1.506 0	0.397 2	1.684 7	0.579 1
		Bagging	0.605 4	2.122 8	0.386 8	1.169 9	0.605 4	0.593 5	1.494 8	0.392 5	1.236 4	0.574 3
		Boosting	0.912 0	1.250 1	0.684 0	0.968 6	0.894 7	0.928 5	1.037 0	0.730 7	0.901 4	0.950 8
无滞后		DI	0.596 0	1.976 6	1.430 9	1.681 0	0.600 7	0.590 3	1.970 2	1.354 4	1.855 4	0.593 5
		Bagging	0.605 4	1.921 6	1.446 7	1.690 4	0.611 7	0.593 5	1.914 4	1.354 4	1.783 6	0.595 1
		Boosting	0.912 0	1.913 7	1.341 3	1.173 1	0.907 3	0.928 5	2.113 8	1.351 3	1.421 4	0.925 3

	方法		移动窗口					累计窗口
	方法		模型1	模型2	模型3	模型4	模型5	模型1	模型2	模型3	模型4	模型5
有滞后		DI	0.596 0	0.550 4	0.404 1	1.328 7	0.520 5	0.590 3	0.449 9	0.473 8	1.054 5	0.504 1
		Bagging	0.605 4	0.566 1	0.459 2	1.383 8	0.536 2	0.593 5	0.480 2	0.488 2	1.021 0	0.499 3
		Boosting	0.912 0	0.789 4	0.652 6	1.163 6	0.795 7	0.928 5	0.721 1	0.727 5	0.907 7	0.877 4
无滞后		DI	0.596 0	1.765 9	1.638 5	2.179 4	0.555 1	0.590 3	2.054 8	1.555 5	2.239 9	0.529 7
		Bagging	0.605 4	1.740 7	1.654 2	2.155 9	0.536 2	0.593 5	2.002 2	1.555 5	2.190 4	0.532 8
		Boosting	0.912 0	1.747 0	1.298 9	1.182 5	0.813 0	0.928 5	1.730 9	1.328 9	1.233 2	0.818 4

Nowcasting of China’s industrial added value based on electric power big data

RichHTML

PDF下载

Knowledge

Abstract

Cite this article

share this article

Figures/Tables 10

References 11

Related Articles 1

Metrics

Recommended 0

	方法		移动窗口					累计窗口
	方法		模型1	模型2	模型3	模型4	模型5	模型1	模型2	模型3	模型4	模型5
有滞后		DI	0.596 0	0.245 3	0.086 5	0.240 6	0.580 2	0.590 3	0.293 5	0.118 1	0.191 4	0.571 1
		Bagging	0.605 4	0.256 3	0.092 8	0.209 1	0.592 8	0.593 5	0.299 9	0.129 2	0.183 5	0.571 1
		Boosting	0.912 0	0.283 0	0.209 1	0.283 0	0.899 5	0.928 5	0.374 9	0.454 7	0.424 4	0.910 9
无滞后		DI	0.596 0	0.506 3	0.449 7	0.493 8	0.589 7	0.590 3	0.433 9	0.406 8	0.422 8	0.583 9
		Bagging	0.605 4	0.504 8	0.452 9	0.489 0	0.599 1	0.593 5	0.433 9	0.406 8	0.429 1	0.585 5
		Boosting	0.912 0	0.613 3	0.597 5	0.608 5	0.905 7	0.928 5	0.646 1	0.599 8	0.654 1	0.923 7