Data stream mining is a challenging task because the data come only one or a chunk at a time. An online learner has to learn while operating continuously. Such a scenario occurs in numerous real-world scenarios, e.g., condition monitoring, fault diagnosis, energy consumption, medical tests, financial information, etc. To make the situation more challenging, the underlying distribution of a data stream may change over time (i.e., a concept drift). This talk first introduces the learning-in-the-model-space framework, which can be used effectively to learn noisy and complex data streams. Online fault diagnosis will be used as an example to illustrate how learning-in-the-model-space could facilitate detecting and classifying unknown faults. Then this talk will present an ensemble approach to tackling concept drifts, i.e., adapting the ensemble diversity after a drift is detected in order to learn new concept quickly and accurately. Finally, this talk will describe a new method for detecting both real and virtual drifts more accurately.
Period
21 Jul 2025 → 23 Jul 2025
Event title
The 9th Euro-China Conference on Intelligent Data Analysis and Applications