Gabriel FERIANCIC

Biographical note
Gabriel Feriancic is a Technical Director at Sistran Engenharia and a professor at Universidade de São Paulo (Brazil). He received his Master of Science degree in Logistics Systems Engineering (2005), and his PhD in Transportation Engineering (2010), both from Polytechnic University of Sao Paulo, Brazil.

Presentation title: CCDR Big Data and Traditional Survey Origin-Destination Matrixes Comparison in a Brazilian Case
The objective of the research was to compare two different Origin-Destination Matrixes from two different sources. The first one, a traditional transport household survey (home-based interviews), conducted by the Public Authority and the other a CDR (call data records) dataset generated from a Big Data analytical platform provided by a mobile carrier corporation.
Pindamonhangaba city was selected among the candidate cities because of its higher reliability OD survey. Pindamonhangaba holds more than 160 thousand habitants and is part of the ‘Vale’ Metropolitan Area, which comprises 39 municipalities with almost two and a half million habitants.
Both matrixes were expanded using the the national census survey population data. The average expansion factor for the traditional survey was 47,9 and mobile phone carrier detained a 25,9% market share. One of the outcomes from the comparison between those different strategies revealed that the CDR approach estimates a larger number of trips than the traditional transport household survey (around 35% higher), when comparing a full work-day.
The temporal travel patterns also differed when comparing the two databases. During the morning rush hour, the CDR dataset has estimates fewer trips than the household survey. Furthermore, using the CDR approach there appears to be almost12% more trips during the evening rush period, when compared to the household survey. This deviation can be explained by a time delay on the use of mobile phones when users reach their workplaces, affecting the precise time identification of trips.
One of the most important finding of this research was the CDR Matrix capability to depict more complete travel patterns; whereas the household survey tends to concentrate trips in the main origin-destination pairs, the CDR showed more complex travel patterns, including shorter, chained trips. This could be considered a result of lack of representativeness of the household survey sample. The CDR Matrix provided reasonable sample between every traffic zone within the area of analysis, generating a more complete representation of travel patterns of the assessed population.
This tool can be used in many kinds of transport problems such as complementation of origin-destination survey data, transport models calibration, travel patterns of different land uses and seasonality studies.