- Data Analytics
As the fuel economy of vehicles increases, both costs and carbon emissions can be reduced. However, understanding what drives the fuel economy of semi-trucks is a challenging feat due to the number of variables involved. This capstone analyzed 21 variables related to the efficiency of semi-trucks to determine how they interact with one another, and furthermore, to quantify and rank their respective impacts on the fuel economy of semi-trucks. To perform this analysis, a dataset comprising these 21 variables across 2,606 semi-trucks over five years of time was used. The nonlinearity of the data necessitated the implementation of machine learning methods, such as K Neighbors Classifier and Random Forest. In addition to this dataset, 12 months of data from an electric semi-truck was studied to understand the current feasibility of this technology. This capstone’s analysis revealed that miles accumulated on tires and vehicle weight have the most significant impact on fuel economy. In finding these results, the K Neighbors Classifier model achieved an accuracy of 64%. The conclusions reached through this research can be utilized by the sponsoring company, and their partners, to improve the fuel economy of their respective fleets. By focusing on these specific variables, the resulting increase in fuel economy will lead to both decreased carbon emissions and expenses on fuel. Additionally, it will lead to the greatest return on investment with their expenditures on vehicle modifications, in terms of dollars spent to miles per gallon realized. The results from the electric vehicle analysis were promising, with the technology showing the potential to increase vehicle efficiency by 35%. This discovery, coupled with the lower cost of electricity compared to diesel fuel, paves an auspicious road for electric semi-trucks if the many current infeasibilities, namely traveling distances possible and long charging times, can be solved.