Electric vehicles are revolutionizing the private and commercial vehicle industries. Electric vehicles can not only been seen as private vehicles but a lot of comemrcial companies such as taxi ride companies and corporations are utilizing electric vehicles for they’re efficiency and environmental benefits. The goal was to determine what factors accelerate the adoption of electric vehicles. Whether electric vehicles are equally popular in each state. Can it be determined what drives the adoption and growth of electric vehicles in a specific region.
Various factors such as GDP, percentage of individuals with a bachelors degree, population of different age groups, energy consumption by state, average temperature by year, median salary by state, price of electricity in each state and total number of vehicles were considered for this analysis. These factors were chosen based on maintaining a balance of economic and social factors that represent a state. It was expected that these factors would provide a somewhat comprehensive state of the US States in the particular year. The data was collected for years 2008-2020. This time frame was chosen becasue the rise of electric vehicles happened in this era. The ending year was mostly forced due to lack of data regarding electric vehicle registrations in the USA. During the analysis it was found that some of the features would turn out to be not so influencial in determine the number of electric vehicles in the state.
It was found from the analysis that California, Hawaii, Utah, Washington, New York, DC, Vermont, Virgina, Arizona and Oregon had the highest percentage of electric vehicles out of the total percentage. Although they had the highest percentages in the data, the highest percentage overall was only 1.1% of total electric vehicles.
It was determined tha tthe factors currently in use to predict the number of registrations in each state were not effective. It was required to look at more focused features that relate to the growth of electric vehicles in each state. The GDP was modified to GDP per working population (adults from 19-54), energy consumption per working population, number of total jobs per working population and also another feature which was number of charging ports per working population was added. Another huge change that was needed to be made was using the year in the analysis. It is clear that in the earlier part of 2008, the percentage of electric vehicles out of total number of vehicles was chosen as the new value to predict instead of merely number of registration. This change was necessary to account for the difference in population and economic factors in each state.
Using Association Rule mining it was found that High percentage of electric vehicles in a state had very strong correlations with High GDP per working population, High percentage of population with a bachelors degree, High median income, High number of charging ports per working population, and Low energy consumption per working population. This was a great discovery which supported the hypothesis that highly educated areas with higher median incomes see a higher percentage of electric vehicles. Although, this hypothesis was not fully validated, since high education does not directly correlate to high GDP.
Applying Machine learning algorithms to train the models and apply them to the testing data is only a small part of this project. The future values of all the features need to be predicted using predictive models such as SARIMA or prophet. Once these future values going out to 2030 are predicted, the best trained model which was Support Vector Machine can be applied on the predicted features to get insight on how the economic, educational and social factors affect the adoption and growth of Electric vehicles in the USA. Another great change to this analysis could be looking at monthly values instead of annual values, to allow the model to have more data to train on.