Time Series prediction with deep learning
Vanessa Haykal is in her third year of PhD in Computer Science and is interested in deep learning methods for time series. The aim is to predict the future values of time series using recent methods in machine learning.
Sequential data processing is an important part of the problems addressed by machine learning. A time series is a series of numerical values representing the evolution of a specific quantity over time. Such series of variables can be expressed mathematically in order to analyse its behaviour and to understand its past evolution and to predict future behaviour. There are different approaches for forecasting. Conventional methods are not always adapted to the characteristics of these temporal data.
Deep learning is based on a network of artificial neurons inspired by the human brain. This network consists of several layers of neurons, each receiving and interpreting information from the previous layer. For example, the system will learn to recognize letters before tackling words in a text, or determine if there is a face on a photo before discovering who they are.
The aim of this thesis is to develop new deep architectures and corresponding learning methods able to tackle difficulties encountered in complex time series. Experiments on financial data as well as electric consumptions are carried out to show the interest of the proposed models.