Avignon: détection de corruption marchés publics

Intitulé du stage / Internship title: Analyse de graphes signés pour détecter la corruption dans les marchés publics / Analysis of signed graphs for the detection of corruption in public procurements Encadrants / Advisors : Rosa Figueiredo Vincent Labatut Lieu du stage / Location: LIA-CERI, Avignon (ceri.univ-avignon.fr ) Internship description Context The advent of public open data is particularly important in the context of public procurements, as recent laws force public institutions to publish their public procurement-related data. Large amounts of data are already available, and soon the activity of this whole sector will be described and ready to be used by anyone. One can already perform analyses which were impossible not long ago, as it was difficult or even impossible to access these data. Processing and analyzing public procurements data is a major academic and societal challenge, due to the considerable amounts at stake (public procurements correspond to approximately 15% of the GDP, i.e. 200 billion Euros a year for France) as well as the economical and societal benefits expected from this imperative of transparency. This internship is related to two themes developed at the [ http://lia.univ-avignon.fr/ | LIA ] (Avignon Computer Science Lab): Complex Systems and Digital Society. It is connected to Computer Science, Economics, Law, and is a part of DéCoMaP (Detection of corruption in public procurements), a larger project funded by the [ https://anr.fr/ | ANR ] (French NSF) and aiming at retrieving, processing and analyzing open data related to French public procurements, in order to design a tool able to assess corruption risks between public buyers and suppliers. In order to obtain reliable models, it is fundamental to take into account the dual nature of interactions between suppliers and buyers (i.e. sane vs. corrupted contractual relation). A good way of doing so is to use signed graphs as models. Unlike unsigned graphs, signed graphs allow representing antagonistic relationships. Moreover, they are much less studied than their unsigned counterparts, so there this domain is very open. Designing automatic tools for the assessment of risks of corruption in public procurements is a task called ?red flagging?. It is not completely new, as some teams have been working on it for a few years [1], especially at the European level [2]. However, none of them applies to French public procurements, as they do not handle its specifics (legal framework, nature and form of the available open data). Moreover, existing approaches focus on individual information, which characterize buyers and suppliers independently, and ignore relational information, which corresponds to interactions and interdependencies between these agents. Consequently, these approaches potentially miss a number of emerging features in the studied system, i.e. properties which exist only at a higher granular level than that of the isolated agent, and which can be essential to fully understand they system. Tasks The work conducted before this internship allowed us to retrieve and store in a local database all the data regarding French public procurements as described by the French official journal (Bulletin Officiel des Annonces des Marchés Publics - [ https://www.boamp.fr/ | BOAMP ] ). We now want to identify the relevant information, and complete them using both state of the art results (legal analysis, theoretical and empirical economical approaches) and human expertise through [ https://cv.archives-ouvertes.fr/pierre-henri-morand | Pierre-Henri Morand ], the economics professor leading the DeCoMaP project. Depending on how the project goes, it can be necessary to enhance the database by leveraging other sources such as [ https://ted.europa.eu/TED/main/HomePage.do | TED ] (European version of the BOAMP). The intern will use signed graphs to model, visualize and analyze the complex networks representing the interactions between companies (suppliers) and public administrations (buyers). These are graphs whose edges are characterized by a sign, which can be either positive or negative, allowing to represent antagonistric relationships. These graphs will be built based on our database. As illustrated by the above Figure, this task can be conducted through several different methods. The intern will have to implement, assess and compare them. The extracted graphs will be used for two distinct purposes. The first is to perform a descriptive analysis through standard tools coming from complex networks analysis (size, density, transitivity, centrality, etc.). The second is to solve various signed graph partitioning problems through the application of existing resolution methods [3,4], as well as versions of these methods that the intern will adapt to the characteristics of our own networks. The obtained partitions will be used to identify groups of agents (buyers and suppliers) likely to be connected by unlawful practices. Perspectives If both the intern and adivsors are satisfied with the way the internship went, the intern will have the opportunity to start a PhD funded by the previously mentioned DéCoMaP ANR project . This PhD will be co- advised by [ https://cv.archives-ouvertes.fr/rosa-figueiredo | Rosa Figueiredo ] and [ https://cv.archives-ouvertes.fr/vlabatut | Vincent Labatut ], as for this internship, and additionally by [ https://perso.univ-st-etienne.fr/largeron/ | Christine Largeron ] ( [ https://laboratoirehubertcurien.univ-st-etienne.fr/ | Laboratoire Hubert Curien ] - [ https://www.univ-st-etienne.fr/ | Université Jean Monnet ] ). References [ 1 ] Fazekas, Mihály, et István János Tóth. [ https://www.u4.no/publications/new-ways-to-measure-institutionalised-grand-corruption-in-public-procurement | New ways to measure institutionalised grand corruption in public procurement ] , U4 Brief, n°9 (2014). [ 2 ] Ferwerda, Joras, et Ioana Deleanu. [ https://dspace.library.uu.nl/handle/1874/309580 | Identifying and Reducing Corruption in Public Procurement in the EU ] . European Commission, OLAF, 2013. [ 3 ] Figueiredo, Rosa et Yuri Frota. [ https://www.sciencedirect.com/science/article/abs/pii/S0377221713010205 | The maximum balanced subgraph of a signed graph: Applications and solution approaches ] . European Journal of Operational Research, 236(2) : 473-487, 2014. [ 4 ] Labatut, Vincent. [ https://www.inderscienceonline.com/doi/abs/10.1504/IJSNM.2015.069776 | Generalized Measures for the Evaluation of Community Detection Methods ] , International Journal of Social Network Mining, n°2(2015):44?63.