Big data for modeling human behavior :applications using mobile phone data

In the last few years, the rise of big data has rapidly revolutionized how people communicate, move or organize events, and even how people think and make decisions. With the rise of the production of these big data, along with the increase of computing power and storage, companies, governments and universities altogether are facing a new challenge, that is, finding efficient methods to leverage those large databases and gain useful insights from the information that can be extracted. Mobile phone datasets provide a good lens through which social behavior can be observed, revealing communities, human mobility, and their evolution with time. In this thesis, we address several questions related to the use of big data, and in particular of mobile phone datasets, to extract useful insights to observe and understand human behavior. Firstly, we address methodological concerns regarding the sampling bias existing in those large datasets that may have effects on the results of the analyses. Secondly, we address the topic of spreading processes, as these datasets can also be used to model and predict the spreading of infectious diseases or information among a population. Finally, we take a more practical approach and show that mobile phone data can be used towards humanitarian purposes, by using them as a proxy for food security indicators in Africa.(FSA - Sciences de l'ingénieur) -- UCL, 2015Accès embargo15/09/2015(FSA - Sciences de l'ingénieur) -- UCL, 2015Anglai