基于SVR的区域交通碳排放预测研究

Forecast Study of Regional Transportation Carbon Emissions Based on SVR

基于STIRPAT模型,选择旅客周转量、货物周转量、人均GDP、机动车保有量、碳排放强度、能源结构和城市化率7项指标作为我国区域交通碳排放影响因素,建立基于支持向量回归机的碳排放预测模型,并以1990—2016年北京市交通碳排放相关数据为基础数据做实例分析.结果表明:训练样本交叉验证均方误差仅为0.008 040,得到参数C和γ的最优值;模型预测值与真实值的拟合回归效果良好,训练集和测试集的相关系数分别为0.984 2和0.995 0,即模型具有良好的学习和推广能力;未来区域交通碳排放增长趋势逐渐变缓,但总量将继续呈上升趋势,社会仍然面临较大的温室气体减排压力.

Based on the STIRPAT model, seven influence factors are selected to build a forecasting model based on the SVR（support vector regression）, including passenger-kilometers, freight ton-kilometers, per capita GDP,vehicle population, carbon intensity, energy structure and urbanization rate. Data of Beijing province from 1900 to 2016 is taken as an example for the case study. The results show that the optimal parameters of C and γ are gotten with the training samples＇ cross-validation mean square error of 0.008 040. With the correlation coefficients of 0.984 2 and 0.995 0, the SVR model had a good ability of learning and promotion. Although under a slowing growth trend of carbon emissions, the total transportation carbon emissions would continue to show an upward trend and the society is still faced with great pressure on the reduction of greenhouse gas emissions.