Overall Objectives
Research Program
Application Domains
Highlights of the Year
New Software and Platforms
New Results
Bilateral Contracts and Grants with Industry
Partnerships and Cooperations
XML PDF e-pub
PDF e-Pub

Section: Bilateral Contracts and Grants with Industry

Bilateral Contracts with Industry

ITRAMI : Interactive Trace Mining

Participant : Alexandre Termier.

ITRAMI is a Nano2017 project. Such projects are designed to support joint research efforts between STMicrolectronics and academic partners in the domain of embedded systems. Alexandre Termier is the PI of this project whose goal is to design novel data mining methods for interactive analysis of execution traces. Such methods aim at considerably reducing the time that STMicroelectronics developers spend at understanding, debugging and profiling applications running on STMicrolectronics chips. The work is done at University Grenoble Alps, in collaboration with LACODAM researchers. Two contractual staff members are working on the project in Grenoble: Willy Ugarte as postdoc, and Soumaya Ben Alouane as engineer.

Hyptser : Hybrid Prediction of Time-Series

Participants : Thomas Guyet, Vincent Lemaire [Orange Labs] , Simon Malinowski [LinkMedia] .

Hyptser is a project funded by the Gaspard Monge Program for Optimisation and Operational Research (PGMO). It is dedicated to the development of innovative methods for predictions in time series. In the field of machine learning, ensemble methods have gained popularity in the last years. These methods combine several algorithms that solve the same task in order to improve the performance of the outcome. Two main families of ensemble methods can be found in the literature : The first family makes use of different models and combine their results a posteriori. The methods Bagging and Boosting are examples of methods in this family [26], [34]. The second family is based on a smart selection of the local algorithms in order to create a global hybrid algorithm. Logistic Model Tree [30] or Extreme Learning Machine Tree [36] are examples of such hybrid algorithms. In this project, starting at the end of 2017 for one year, we envision to explore the second family of methods in order to analyze how efficiently hybrid models can perform on the task of time series prediction. We plan to apply these methods to predict resource usage for cloud computing (CPU, memory) so as to minimize their infrastructure.

Particular Contract of the Strategic Action EDF/Inria

Participants : Manel Boumghar [EDF R&D] , Laurent Pierre [EDF R&D] , Thomas Guyet, René Quiniou.

The analysis of customer pathways has become a strategic issue for many businesses. The interaction traces left by clients when connecting to the customer services can be combined with data from other communication channels (phone, web form, e-mail, mail, fax, SMS, shop, etc.) and allow to analyse the customer pathways in details.

Pattern mining tools are able to extract the frequent customer behaviors in very large databases of client pathways. Nevertheless, taking into account the duration and the delay between the customer actions in the mining remains a challenge. The objective of this one-year contract was to design and develop a frequent mining tool that accounts for temporal patterns with negations for analysis of multichannel customer pathways. In this line, we developed and implemented the NTGSP algorithm [17].