How to avoid the black-box effect of artificial intelligence ?

Share Article

Adrien Rosen and Pierre Artoisenet, Senior Consultants in our Energy team gave a detailed presentation at E-world 2020 about how to avoid the black-box effect of Artificial Intelligence in time series forecasting.

Increasing penetration of intermittent generation leads to increasing market volatility and higher risk in the portfolios. The N-SIDE electricity price forecasts help portfolio managers and traders to minimize their risk exposure and embrace new sources of revenues.

However, the complexity inherent to a highly-multivariate and non-parametric regression often lead to the symptomatic Artificial Intelligence “black-box effect”.

N-SIDE has developed a unique expertise in Machine Learning that enables our experts to control the behavior of its models. N-SIDE is bringing together the business and the mathematical knowledge of its experts to deliver the best-in class forecasting models.

The value generation of our end users is at the heart of the developments we do in forecasting. We do not focus on the traditional statistical metrics but on the €/MWh our end-users can generate.

In practice, we systematically go through a series of steps when facing a new forecasting challenge:

  • Which Machine-Learning technology should we use to learn from the past?
  • How can we translate the business objective into mathematical formulation?
  • How do we make sure to fully understand the key drivers in the output of the algorithm?
  • How can we value the business knowledge of our experts and optimally build the inputs that are passed to the algorithm?
  • How to make sure we react fast enough to market changes?
  • How to control the risk of error in the output of the algorithm?

Would you like to receive the slides of the presentation ?

 

Contact Us