This work develops predictive models based on Recurrent Neural Networks (RNNs) and Long Short-Term Memory (LSTM) to forecast pollutant emissions and fuel consumption in vehicles, using historical data such as odometer readings and fuel type. RNNs and LSTMs, due to their ability to capture complex patterns in time series, are applied to identify trends and predict future behaviors, contributing to energy efficiency and emission reduction. Additionally, a multivariate statistical analysis with Random Forest and other algorithms, such as AdaBoost and Gradient Boost, is performed to assess the influence of independent variables on the target variable, identifying critical factors that impact vehicle performance. The combination of these machine learning and data science techniques provides robust and innovative solutions, promoting sustainable development by addressing environmental and economic challenges related to air pollution and the optimization of the transportation sector.