当前位置：网站首页>R language multiple linear regression, ARIMA analysis of the impact of different candidates in the United States on the economic GDP time series

R language multiple linear regression, ARIMA analysis of the impact of different candidates in the United States on the economic GDP time series

2022-08-11 00:50:00 【Rio tinto client laboratory】

全文链接：http://tecdat.cn/?p=28144

原文出处：拓端数据部落公众号

作者：Yuanchang Luo

近段时间,The U.S. presidential election has drawn the attention of countries around the world.Republican candidate Donald·Trump and Democratic candidate Joe·Biden will run for president.Bipartisan candidates in financial trade、Key development areas such as economic and financial governance and prevention and control measures have different political positions and policies.Candidates with different political positions have a certain impact on the economic development of the United States and even the world,This article combines specific data,Quantitatively analyze the impact of different candidates on the development of the U.S. economy.

解决方案

任务/目标

According to the data of various indicators in the United States and the difference between the two candidates 政策,Analyze the impact on the U.S. economy.

数据源准备

Search data on US government public datasets,共 26 类, 并且用 GDP to reflect the U.S. economy.得到数据后, Because it is time series data,Hence by Lagrangian interpolation method to fill empty values.并且,Because it is the number published by the U.S. government 据集,Hence the default outlier,That is, larger and smaller values are determined by real due to historical factors,不作处理.

特征抽取

First, observe the correlation matrix between features and the phase with the dependent variable 关性,Preliminary removal 7 A very low correlation with the dependent variable and an indicator indicators with high correlation,使用剩余１９as an independent variable Metrics for regression analysis.（Dropped indicators：'US personal income 中位数','个人所得税（最高）','个人所得税（最低） ','Export of goods and services','Net acquisition of financial assets','State of the labor market condition index','失业率'）

建模

多元线性回归,General application with multiple characteristic indicators return the problem. in the process of multiple linear regression,In addition to considering the model AIC minimum outer,It is also necessary to consider the relationship between the independent variables among the models the effect on the dependent variable,即多重共线性,通过 VIF to remove relevant independent variables. ARIMA,It is generally used in the field of time series. ARIMA A model refers to the transformation of a non-stationary time series into a flat stable time series,然后将结果变量做自回归（AR）和自平移（MA）.

模型优化

1.通过 VIF The criterion excludes relevant independent variables：

上图为 VIF Initial and final results.进一步筛选 7 indicators to predict.

2.通过 AIC Criteria to choose the optimal model

a combination of factors t test and model AIC,通过向 The method of forward-backward selection,Choose the best regression model.

结合上图,Be confident that this regression model is performing well.

3.Time series forecasting independent variables

Due to the data released by the US government as of now 19 年,而我们需要 21 年 1 The data for the month independent variable is predicted by times The economic impact of the election of different candidates,Hence the passage of time 序列对 5 future forecasts 5 quarterly value.

4.Quantitative effects of different policies on characteristics

Combine policies from different candidates,Can be analyzed qualitatively Indicate whether the effect on each feature increases or decreases,然后通过平均 20 年的数据,Calculate the increase and decrease of each characteristic percent mean,In this way, the specific value of the impact can be estimated. 这样一来,Pass on historical data ARIMA 模型得到 20 year's data,Then through the different politics of each candidate The impact of the strategy on the indicator and the average historical change,就得到了 21 The specific values of each indicator in the four quarters of the year,然后通过 Equations obtained from multiple regression,预测 21 年 4 个季度的 GDP 具体数值.

项目结果

多元回归方程：ｙ＝ − ０．３４７８ − ０．０８５４８ｘ２＋１．５７９ × １０ −７ｘ１０＋４．６５３ × １０ −５ｘ１４＋１．５６５ × １０ −５ｘ１５＋１．１５６ｘ１９

Combine the predicted values for each indicator,Calculate when the different candidates The impact of elections on the economy：

可以看到,The election of both candidates will have a certain boost to the U.S. economy,But Biden's election will undoubtedly improve even more, 因此可以估计,Biden has a better chance of winning this major 选.Evaluating effects cannot be based solely on economic impact,要综合考虑, It is necessary to refer to the impact of the specific policies of different candidates, And the differences between the two candidates in different parties and groups 态度.Therefore, the forecast results are only for reference.