This book presents cutting-edge, effective methods of artificial intelligence and machine learning for data modeling, with applications to the important and challenging area of air pollution. Both classical stochastic and tree-based ensemble learning approaches and their hybrid combinations are considered, including: ARIMA, Temporal Causal Modeling, wavelet transforms, Classification and Regression Trees (CART), Multivariate Adaptive Regression Splines (MARS), Random Forests, and Adaptively Resampling and Combining (Arcing). The selected methods require minimal computer resources (execution time and memory) and are oriented for inclusion in distributed environments and mobile devices. In addition, the book emphasizes the statistically correct construction and investigation of the problems under consideration and detailed analysis of model errors, rather than presenting statistical theory or ready-made codes from software packages. The indicated approaches have been demonstrated to forecast time series of air pollutants such as particulate matter, sulfur dioxide, nitrogen dioxide, and others, depending on a small number of rapidly changing meteorological and atmospheric factors. The methods and frameworks are applied to empirical data from several cities in Bulgaria. The results of the individual applications are presented in five chapters as case studies. These studies demonstrate in detail the steps of the developed approaches for modeling and forecasting of real measured data related to urban air pollution. The book has the potential to serve not only as a systematic introduction to the selected ensemble learning methods for time series, but also as a tool and guide for building adequate and statistically valid forecasting models.