Download PDF
Luca Frare’s thesis work falls within the innovative domain of Big Data applications in transport planning, with a specific focus on constructing Origin-Destination (OD) matrices from cellular network data.
Transport planning and modeling require an accurate estimation of mobility demand (typically represented via Origin-Destination matrices). Traditional demand estimation methods—mainly roadside counts and travel surveys—suffer from high costs, lengthy acquisition periods, and low update frequencies. Consequently, demand reconstruction methodologies based on Big Data have gained significant traction in recent years.
Developed alongside a real-world case study, this thesis project primarily addresses OD matrices derived from mobile network data. Following a concise overview of traditional matrix estimation methods and their associated limitations, the paper outlines the characteristics of mobile phone data, the state-of-the-art methods for reconstructing matrices from these sources, and the inherent constraints of using this data type.
Subsequently, a method is proposed to improve the quality of a mobile-data-derived OD matrix by incorporating field traffic counts and other data available in literature. This methodology is then applied to a case study: the regional traffic model of the Friuli-Venezia Giulia region.
The proposed methodology consists of the following key phases:
- Analysis and processing of mobile phone data
- Collection of field survey data and secondary data sources
- Reconstruction and calibration of the supply network graph
- Matrix calibration using the Ordinary Least Squares (OLS) method to align the matrix with observed field flows
- Validation using statistical analysis tools:
- GEH statistic to evaluate calibration quality
- Correlation analysis against additional available datasets (e.g., ISTAT) for further verification
The conclusions highlight several key advantages of this methodology and data source:
- Cost and Time Efficiency: Compared to traditional sample surveys, the required time and financial investment are substantially reduced.
- High Reliability: Combining mobile network data with field surveys and secondary datasets yields a matrix that closely mirrors real-world conditions, overcoming several limitations inherent to unrefined mobile data.
- Broad Applicability: The proposed methodology enables infrastructure operators, public administrations at all territorial levels, and transport planners to obtain robust data for traffic simulation. This allows for the development of reliable, replicable Decision Support Systems (DSS) to guide both technical and policy decisions.
