Systematic Evaluation of the Evolution of Traffic Flow Prediction and Correlation Analysis Methods

The evaluation of spatial and temporal variability of traffic, as well as its forecasting, is a long-standing topic which has attracted great attention over the years due to the impacts that transportation systems have on the environment and quality of life. The topic, over the years, has expanded in light of new technologies and new methods available. The aim of this work is to identify, on the basis of the literature, the methodologies used for the analysis of traffic variables (flow, speed, etc.), as well as how they have evolved over the years. The analysis is based on a scientific mapping approach, using Bibliometrix as a reference tool. The tool was used to extract information such as the most productive sources, the main authors and the production trends of the sources over time. Among the different approaches for the identification of relevant topics, we used the network approach, which combines co-occurrence networks, thematic maps and thematic evolution to identify methodologies consolidated in the literature, as well as emerging methodologies. Based on this, it was possible to trace the temporal evolution of the topic, and to make a classification of the typologies of methods employed over time (statistical, mathematical, machine learning and deep learning) for the analysis of spatial, temporal and spatio-temporal correlation. As a further outcome, our work brings to light how the topic of spatio-temporal correlation analysis and forecasting of traffic variables is addressed today.