Discovery Challenge ECML PKDD 2024
Discovery Challenge by The European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases (ECML PKDD) is one of Europe’s premier conferences on machine learning and data mining. In collaboration with Volvo Group Truck Technologies, Halmstad University challenges participants to predict the risk of failure for a undisclosed component of Volvo trucks.
Developing a data-driven model for predicting failure risk levels for a component of Volvo heavy-duty trucks
This industrial challenge is organised in collaboration with Volvo Group Truck Technologies. The participants will work with a real-world dataset incorporating the measurements from a fleet of more than 20,000 Volvo heavy-duty trucks. The measurements are collected for a specific period through telemetries or workshop visits. with survival information about an undisclosed component of the trucks will be released for the first time in connection with the challenge.
Who can participate?
Anyone with background in AI, data science, machine learning, predictive maintenance can participate. It can be researchers, practitioners, students etcetera. You participate individually or in a team. The challenge is in English.
Why should you participate?
- Real-World Impact: Your solution could contribute to enhancing the predictive maintenance strategies for trucks, leading to improved reliability and efficiency and reducing the environmental impact and CO2 emissions.
- Networking Opportunities: Foster collaborations and knowledge exchange by meeting with industry professionals, researchers, and fellow participants at ECML PKDD 2024.
- Prizes: Stand a chance to win a free registration for the conference and recognition for your innovative solutions. Additional monetary prizes are expected and will be announced soon.
How do you participate?
Participants create an account on www.codabench.org
External link. and should submit their results (i.e., prediction files) before the deadline to a page on codabench that will be announced later.
Date, time and place
- Place: Online
- Start: 15 May
- End: 15 June
- Results will be published: 29 June
- System description report is submitted: 15 July
More information about the challange
Problem description
The participant will develop a model to predict the risk levels to help Volvo find the components requiring maintenance. This will enable the company to monitor the equipment more intelligently and prevent breakdowns by taking proper actions. The three target classes are low, medium, and high-risk groups.
Dataset
During the challange, a dataset will be released for the first time comprising sensor measurements from over 20,000 Volvo trucks spanning a significant timeframe. This invaluable dataset also includes survival data pertaining to a disclosed component of the trucks.
Tasks
The provided dataset contains information about two generations of the same component. The validation and test data will be shared separately for each generation, but the training data only includes one generation. Accordingly, two tasks are defined. The first task is to design a model that can accurately predict the risk level of a component from the same generation as the provided training data. The second task is to create a model that performs well on the second generation. The main goal of the second task is to assess the model’s generalization ability in new settings without explicitly providing training information. It is also possible to submit the same model for both tasks.
