Fuel Consumption Prediction of Vehicles Using Machine Learning Algorithm

Abstract

The major objective of this study revolves around accurately estimating the distance a car can travel in kilometres per Litter of fuel consumed. This study develops a precise machine-learning algorithm for predicting vehicle petroleum consumption. This encompasses adapting distinct machine-learning techniques, evaluating their performance, selecting the most optimal model, and validating its real-world applicability. The dataset used for this study includes attributes such as Miles per gallon (Mpg), acceleration, horsepower, displacement, cylinder count, and car model. The implementation strategy entails comprehensive data pre-processing and employing well-established machine learning techniques: Random Forest, Decision Tree, and Linear Regression. The Python programming environment is applied for coding and data manipulation. Model performance assessment uses the Mean Squared Error (MSE) metric. The findings show the performance of the Random Forest algorithm as having the lowest MSE value of 0.008806 among the assessed models. In conclusion, the proficiency of the Random Forest algorithm. in predicting fuel consumption will open avenues for informed decision-making and resource optimization within the automotive sector.