This work presents an investigation of tailoring Network Representation Learning (NRL) for an application in the Financial Industry. NRL approaches are data-driven models that learn how to encode graph structures into low-dimensional vector spaces, which can be further exploited by downstream Machine Learning applications. They can potentially bring a lot of benefits in the Financial Industry since they extract in an automatic way features that can provide useful input regarding graph structures, called embeddings. Financial transactions can be represented as a network, and through NRL, it is possible to extract embeddings that reflect the intrinsic inter-connected nature of economic relationships. Such embeddings can be used for several purposes, among which Anomaly Detection to fight financial crime.This work provides a qualitative analysis over state-of-the-art NRL models, which identifies Graph Convolutional Network (ConvGNN) as the most suitable category of approaches for Financial Industry but with a certain need for further improvement. Financial Industry poses additional challenges when modelling a NRL solution. Despite the need of having a scalable solution to handle real-world graph with considerable dimensions, it is necessary to take into consideration several characteristics: transactions graphs are inherently dynamic since every day new transactions are executed and nodes can be heterogeneous. Besides, everything is further complicated by the need to have updated information in (near) real-time due to the sensitivity of the application domain. For these reasons, GraphSAGE has been considered as a base for the experiments, which is an inductive ConvGNN model. Two variants of GraphSAGE are presented: a dynamic variant whose weights evolve accordingly with the input sequence of graph snapshots, and a variant specifically meant to handle bipartite graphs. These variants have been evaluated by applying them to real-world data and leveraging the generated embeddings to perform Anomaly Detection. The experiments demonstrate that leveraging these variants leads toimagecomparable results with other state-of-the-art approaches, but having the advantage of being suitable to handle real-world financial data sets. / Detta arbete presenterar en undersökning av tillämpningar av Network Representation Learning (NRL) inom den finansiella industrin. Metoder inom NRL möjliggör datadriven kondensering av grafstrukturer till lågdimensionella och lätthanterliga vektorer.Dessa vektorer kan sedan användas i andra maskininlärningsuppgifter. Närmare bestämt, kan metoder inom NRL underlätta hantering av och informantionsutvinning ur beräkningsintensiva och storskaliga grafer inom den finansiella sektorn, till exempel avvikelsehantering bland finansiella transaktioner. Arbetet med data av denna typ försvåras av det faktum att transaktionsgrafer är dynamiska och i konstant förändring. Utöver detta kan noderna, dvs transaktionspunkterna, vara vitt skilda eller med andra ord härstamma från olika fördelningar.I detta arbete har Graph Convolutional Network (ConvGNN) ansetts till den mest lämpliga lösningen för nämnda tillämpningar riktade mot upptäckt av avvikelser i transaktioner. GraphSAGE har använts som utgångspunkt för experimenten i två olika varianter: en dynamisk version där vikterna uppdateras allteftersom nya transaktionssekvenser matas in, och en variant avsedd särskilt för bipartita (tvådelade) grafer. Dessa varianter har utvärderats genom användning av faktiska datamängder med avvikelsehantering som slutmål.

No description available.

Ces travaux sont du domaine de génie systèmes industriels et l'ingénierie de l'innovation. Ils se sont déroulés dans l'industrie de services financiers au sein du groupe de banque-assurance Groupama. Les services représentent 64% de PIB mondial. Le secteur employait en 2007 en France près de 20 millions de personnes contre 5 millions pour l'industrie. Pourtant la recherche sur les processus d'innovation présente un déséquilibre avec un article scientifique NSD (new service development) pour quatre NPD (new product development). L'objectif de ces travaux est de proposer une méthode de pilotage des projets d'innovation dans les services. Pour formaliser le pilotage d'innovation dans les services tout en préservant la zone de liberté nécessaire à l'innovation, nous sommes basés sur un formalisme de type NPD pour définir un processus NSD qui respecte la flexibilité spécifique de l'innovation dans les services. Le coeur de notre hypothèse a été d'identifier des invariants de processus NSD. Nous proposons six classes d'invariants : les OICs (Objets Intermédiaires de Conception), les ressources, les compétences, les tâches, les indicateurs et les méthodes. Leurs interactions sont rendues dans un Diagramme de Classes UML. Un projet peut être décrit comme une « somme » d'OIC eux-mêmes résultat de l'agencement des 5 autres invariants. Ces six classes d'invariants ont été validées sur projets de Groupama. Elles permettent de décrire, suivre, capitaliser, réutiliser des savoirs acquis sur des projets antérieurs et de manager les projets innovants. Nous proposons aussi un processus de pilotage des projets NSD, formalisé par un Diagramme d'Ordonnancement des Phases sous MEGA / The present thesis is about the field of system engineering and innovation engineering. It took place in the financial industry within the group of banking insurance Groupama. Services represent 64% of world's GDP (Gross Domestic Product). This branch employed in 2007 in France about 20 million people against 5 million for industry. Still the research on the innovation processes presents a gap with only one NSD (new services development) article against four NPD (new products development). The aim of this work is to propose a method for management of innovation projects in the services branch. To formalize the management of innovation projects for the services, while preserving the space required for innovation, we use a NPD's type formalism in order to define a NSD process while respecting the specific flexibility of the innovation in the services branch. The center of our hypothesis was to identify invariants within the NSD process. We propose six classes of invariants: IDOs (Intermediate Design Objects), the resources, the skills, the tasks, the indicators and the methods. Theirs interactions are shown in a UML Diagram of Classes. A project may be represented as an "addition" of IDOs, where they are the result of 5 other invariants arrangements. These six classes of invariants were validated on Groupama projects. They are relevant to describe, monitoring, capitalize, re-use of the knowledge acquired on previous projects and to manage innovative projects as well. We also propose a process of piloting of the NSD projects, formalized by a Diagram of Phases Sequencing of MEGA