Section: Application Domains

Industry

We present below our industrial collaborations. Some are well established partnerships, while others are more recent collaborations with local industries that wish to reinforce their Data Science R&D with us (e.g. STMicroelectronics, Energiency, Amossys).

• Execution trace analysis for SOC debugging (STMicroelectronics). We have an ongoing collaborations with STMicroelectronics, which is one of the world top-5 electronic chip makers. Nowadays, set-top boxes, smartphones or onboard car computers are powered by highly integrated chips called System-on-Chip (SoC). Such chips contain on a single die processing units, memories, IO units and specialized accelerators (such as audio and video encoding/decoding). Programming SoC is a hard task due to their inherent parallelism, leading to subtle bugs when several components do not deliver their results within a given time frame. Existing debuggers and profilers are ill-adapted in this case because of their high intrusivity that modifies the timings. Hence the most used technique is to capture a trace of the execution and analyze it post-mortem. While Alexandre Termier was in Grenoble he initiated several works for analyzing such traces with pattern mining techniques, which he is now pursuing with his colleagues of the Lacodam project-team.

• Resource consumption analysis for optimizing energy consumption and practices in industrial factories (Energiency). In order to increase their benefits, companies introduce more and more sensors in their factories. Thus, the resource (electricity, water, etc.) consumption of engines, workshops or factories are recorded in the form of times series or temporal sequences. The person who is in charge of resource consumption optimization needs better software than classical spreadsheets. He/she needs effective decision-aiding tools with statistical and artificial intelligence knowledge. The start-up Energiency aims at designing and offering such pieces of software for analyzing energy consumption. The starting CIFRE PhD thesis of Maël Guillemé aims at proposing new approaches and solutions from the data mining field to tackle this issue.

• Security (Amossys). Current networks are faced with an increasing variety of attacks, from the classic « DDoS » that makes a server unusuable for a few fours, to advanced attacks that silently infiltrate a network and exfiltrate sensitive information monthes or even years later. Such intrusions, called APT (Advanced Persistent Threat) are extremely hard to detect, and this will become even harder as most communications will be encrypted. A promising solution is to work on “behavioral analysis”, by discovering patterns based on the metadata of IP-packets. Such patterns can relate to an unusual sequencing of events, or to an unusual communication graph. Finding such complex patterns over a large volume of streaming data requires to revisit existing stream mining algorithms to dramatically improve their throughput, while guaranteeing a manageable false positive rate. We are collaborating on this topic with the Amossys company and the Emsec team of Irisa through the co-supervision of a CIFRE PhD (located in the Emsec team). Our goal is to design novel anomaly detection methods that can detect APT, and that scales on real traffic volumes.

• Market basket data analysis (Intermarché) and multi-channel interaction data analysis (EDF) for better Customer Relationship Management (CRM). An important application domain of data mining for companies that deal with large numbers of customers is to analyze customer interaction data, either for marketing purposes or to improve the quality of service. We have activities in both settings. In the first case, we collaborate with a major french retailer, Intermarché, in order to detect customer churn by analyzing market basket data. In the second case, we collaborate with the major french power supplier, EDF, to discover actionable patterns for CRM aiming at avoiding reaching undesirable situations from logs of user interactions with the company (web clicks, phone calls, etc.).