International Choice Modelling Conference, International Choice Modelling Conference 2015

Font Size: 
INCORPORATING SOCIAL MEDIA DATA IN TRAVEL CHOICE MODELS: CONCEPTUAL FRAMEWORK AND EXPLORATORY ANALYSIS
Ying Chen, Andreas Frei, Hani S Mahmassani

Last modified: 18 May 2015

Abstract


In applying travel choice models to predict demand in practice, it is generally assumed that the choice set is known and static, that travelers are aware of the attributes of the available alternatives, and that the individual choices satisfy the needs/preferences of the decision-maker viewed independently (of others). This simplification of the underlying behavioral processes becomes less acceptable for many applications, especially when modeling choices with potentially large choice sets, attributes that are perceived differently by each user and choices where decision-makers interact heavily with each other.  Such compromises have been adopted mainly because of a lack of data and also because of computational and theoretical convenience in transportation and econometric models. New data sources and modern survey technologies have potential to overcome the lack of detail in data, as well as deliver high levels of spatial and temporal specificity and accuracy. However, they also give rise to new challenges that must be overcome in order to make them usable for real world applications.

This paper discusses the opportunities as well as some of the challenges of incorporating social media data in choice models, illustrated by analyzing the data from a popular on-line location-based social network called Brightkite (1). The data set contains dynamic user check-in information with time stamp and coordinates as well as an undirected social network of its users. The premise of this paper is that social networks influence people’s resource allocation decisions in planning and executing activities, and that the manifestation of such decisions in their travel patterns can be observed in the data set at hand. Location-based social networking data provide an important new dimension in understanding travel choice behavior, by providing high levels of location and time accuracy over longer time frames in conjunction with explicit friendship network information. Such data allow for studying location choice dynamics and social networking aspects explicitly.  However, such data can also suffer from well-known biases, due to the sporadic nature of check-ins (which leads to non-random censoring), inferring the choice set, noisy data and self-selection.

Furthermore, with the rapid diffusion of information and frequent social communication taking place through various channels nowadays, not all travelers are aware of the same information, and they likely perceive information differently depending on the reliability, latency and affinity of the source. Travelers learn about changes through various information sources, including social communication with others. For example, the formation of destination choice sets will be influenced not only by external factors but also personal perceptions. In these cases, consideration of the opinion formation process about a new destination and the role of information diffusion through social communication become important elements of the dynamics of the choice set adoption and formation process.  The interpersonal mechanisms behind opinion formation, such as word-of-mouth (2), mass-media (3), direct experience (4) and belief learning (5), may correlate with information exchange. Thus, exploring the choice set generation process also requires investigating the dynamics of social networks and repeated behavior. One would expect that people exhibit strong periodic behavior in traveling between their homes and workplaces, and also in visiting leisure locations.  Moreover, daily travel behavior patterns, especially non-frequent movements such as long distance trips, may be shaped by social networks, that influence the probability positively to visit locations close to our friends, or locations that were visited by our social network before us.

From a preliminary analysis of the Brightkite data, it can be found that friendship status influenced people’s destination choices, especially in a denser social network. Travelers are more likely to visit places that their friends have previously visited, all else being equal. However, the influence of one’s experience on the likelihood of returning to previously visited locations is even stronger. This indicates that the opinion of others matters, but not as much as one’s own experience.

 

KEY WORDS: Social network, Check-ins, Choice set generation, Travel behavior dynamics

REFERENCES

1.     http://en.wikipedia.org/wiki/Brightkite.

2.     Gladwell, M. The Tipping Point: How Little Things Can Make a Difference, Little and Brown, New York, 2002.

3.     Leskovec, J., L.A. Adamic, and B.A. Huberman. The Dynamics of Viral Marketing. Journal of the Association for Computing Machinery, 2006, pp.1-28.

4.     Brenner, T. Decision Making and the Exchange of Information. Book chapter in Self-Organization of Complex Structures: From Individual to Collective Dynamics, 1997, pp.379-392.

5.     Chen, R. and H.S. Mahmassani. Travel Time Perception and Learning Mechanisms in Traffic Networks. In Transportation Research Record: Journal of the Transportation Research Board, No. 1894, Transportation Research Board of the National Academies, Washington, D.C., 2004.


Conference registration is required in order to view papers.