摘要
新能源汽车的普及,有利于减少大气污染,提高空气质量。但与新能源汽车相配套的公共充电基础设施、维修服务等问题却阻碍了新能源汽车销量的增长。因此,预测我国新能源汽车销量以完善相关配套措施、促进新能源汽车产业的发展就显得尤为重要。针对新能源汽车产业属于新兴产业,其相关历史数据较少,且销量变动较大以及影响其销量的因素存在非线性关系的特点,本文利用鲁棒性强的支持向量回归,以及具有较强的抗噪声能力的Bagging集成学习方法,对我国新能源汽车的销量进行预测和分析。首先,选取影响消费者购买意愿的公共充电桩数量和决定消费者购买能力的居民可支配收入作为模型的自变量,并收集相关数据;其次,从原始样本中随机抽取样本量为20的5个相互独立的样本集,并使用6个训练数据对这5个样本集进行训练,得到5个支持向量回归模型;然后,平均5个模型的结果,减少模型噪声,优化最终预测效果;最后,分析所得的预测新能源汽车销量模型的准确性及不足之处。
The popularization of new energy vehicles is conducive to reducing air pollution and improving air quality.However,problems such as public charging infrastructure and maintenance services supporting new energy vehicles have hindered the growth of new energy vehicle sales.Therefore,it is particularly important to predict the sales of new energy vehicles to improve related supporting measures and promote the development of the new energy vehicle industry.Based on the characteristics of the new energy automobile industry,its relatively small related historical data,the fl uctuating sales volume,and the factors that aff ect its sales volume have non-linear relationships,this article uses robust support vector regression and the bagging integrated learning method of anti-noise ability,predicts and analyzes the sales of new energy vehicles.First,the article selects the number of public charging piles that aff ect consumers'purchasing intentions and the disposable income of residents that determine consumers'purchasing power as independent variables of the model,and collects relevant data;second,the article randomly selects 5 samples with a sample size of 20 from the original separate sample sets,and uses 6 training data to train these 5 sample sets to obtain 5 support vector regression models;then,the article averages the results of the 5 models to reduce model noise and optimizes the fi nal prediction eff ect;fi nally,the article analyzes the accuracy and defi ciencies of the resulting model for forecasting sales of new energy vehicles.
出处
《时代汽车》
2021年第10期62-63,共2页
Auto Time
关键词
支持向量回归
集成学习
新能源汽车
居民可支配收入
公共充电桩数量
support vector regression
ensemble learning
new energy vehicles
residents'disposable income
number of public charging piles