Mohamed Quafafou is a professor of computer science at Aix-Marseille University. His main research interests are in Data Mining theory and applications.  Prior to joining Aix-Marseille University at 2005, he served as Professor at Avignon University from 2002 to 2004 and as Assistant Professor at University of Nantes from 1995 to 2001. Mohamed Quafafou received his Habilitation Ph.D. in 2000 on Rough sets Theory and Data mining and a Ph.D. in Computer Science from INSA Lyon, in 1992. He received a M.Sc. in computer science from Paul Sabatier University (Toulouse), a M.Sc. in mathematics from Pau University and the B.Sc. degree in Mathematics from the Cadi Ayyad University (Marrakech).

For about 20 years, he developed research on Rough Set Theory, data mining, web information extraction, etc. In partnership with France Telecom, he developed one of the first web mining systems, whichis dedicated to the analysis of French web to discoveremergent web communities. He headed the Geobs Data Analyzer project which aimed to provide a theoretical foundation for spatial data mining systems covering a large area of applications, e.g., environment, marketing, social analysis, etc. He is currently leading research on data mining especially when data are generated by human activities. His research has a broader scope including personalization and recommendation, data science, web services; social network analysis, etc.He has supervised more than 15 Ph.D. thesis and more than 30 M.Sc. student research projects. He has been expert for a number of French National and International projects.

He is a teacher at Polytech'Marseille School, where he leads the Information System Engineering. He has also other teaching activities in M.Sc level in both Aix-Marseille University and University of Science and Technology of Hanoi. His website ishttp://www.quafafou.com

Nowdays, different actors would like to take advantage of the ubiquitous access to the Internet to provide new and innovative services. They are also interested in evaluating the impact of such pervasive network access on urban services and businesses: transport services, tourist offices and sites, restaurants, cinemas, stores, etc. The emergence of Big Data is an attempt to answer questions of the modern society by allowing the development of ecosystems containing different storage databases, computing resources, methods and algorithms to deal with different kind of data.We advocate that bringing such ecosystems to the vicinity of the users by applying urban sensing methodologies is crucial and of high impact in today's society where smartphones are being increasingly used for accessing the Internet.

This talk presents two problems related to analysis of data, which is generated by human, in the context of smart cities. The first concerns trajectories mining and patterns visualization in order to understand human behavior in the city.The second concerns supervised learning from citizens, where we do not have control over the quality of each labeler. In this case,some annotators may provide random or bad quality label, while others may have good intentions and label the data set given seriously.In order to face this learning from crowds problem, we introduce a Bayesian approach considering data multi labeling, data quality and classes of annotators (experts, spammers, etc.).

The talk concludes with a new line of research that makes a bridge between perception and set theories in order to extend data mining to take into accountnot only the precision and uncertainty of humans but alsotheir perceptions.

Key words:Smart city, human behavior, Patterns extraction, Formal concept analysis, Spatio-Temporal Granularity, Bayesian models, Trajectory, Uncertainty and Imprecision.