This research concerns the use of machine learning techniques for extracting relevant information from real massive datasets. A particular attention is devoted to techniques of feature selection, causal inference, model selection and validation and long-term prediction.
- Long term prediction of time series: reliable and accurate prediction of time series over large future horizons has become the new frontier of the forecasting discipline. Current approaches to long-term time series forecasting rely either on iterated predictors or direct predictors. We proposed a multi-output extension of our previous work on Lazy Learning and we showed that this prediction strategy can be particularly effective in multiple-step-ahead tasks.
- Modelling and distributed compression of wireless sensor data: Wireless sensor networks form an emerging class of computing devices capable of observing the world with an unprecedented resolution, and promise to provide a revolutionary instrument for environmental monitoring. In environmental monitoring studies, many applications are expected to run unattended for months or years. Sensor nodes are however constrained by limited resources, particularly in terms of energy. We proposed a machine learning approach which combines time series prediction and model selection for reducing the amount of communication. The rationale of this approach, called adaptive model selection, is to let the sensors determine in an automated manner a prediction model that does not only ﬁts their measurements, but that also reduces the amount of transmitted data. Secondly we designed a distributed approach for modeling sensed data, based on the principal component analysis (PCA).
- Application of data mining to several domains: motion analysis, anesthesiology, indoor localization and tracking, cryptography (side-channel attack), fraud detection, power systems and remote sensing.