Research On The Development Of New Energy Vehicles Based On Web Search Data And Machine Learning Methods

Nowadays,the world energy situation is grim and the environmental protection awareness is increasing,the new energy vehicles(NEVs)has become the emerging industry.An objective and accurate analysis and forecast of NEVs sales will not only help the government to formulate new macro-control policies,but also help NEVs enterprises to determine their competitive advantages and formulate reasonable marketing strategies.With the continuous development of network technology,the web search data containing a lot of consumer behavior information provides a new way to study the development of NEVs.This paper analyzes BYD’s sales of NEVs and Tesla(China)Model 3 based on web search data.Firstly,a system scheme for constructing a web search keyword database is proposed.Secondly,grey correlation analysis and random forest importance ranking are used to extract characteristic variables,common variables are defined as the third type of characteristic variables,and the main influencing factors are found,which are policies,supporting facilities,private brand building,competitive products,car website,and based on these pertinent suggestions for the development of the NEVs industry were put forward.Finally,traditional SVR is improved by feature weighting,GRAFW-SVR and RFFWSVR are obtained,and RF,KNN,SVR,GRAFW-SVR,RFFW-SVR are used to build predictive models.The results show that the two sets of feature extraction methods can improve the model accuracy.For BYD’s NEVs,the RFFW-SVR-GRA model has the best effect,while for Tesla(China)Model 3,the RFFW-SVR-RF model has better effect;At the same time,sorting according to the R2 of the prediction model,we can get RFFWSVR>GRAFW-SVR>SVR.