Select Page

Machine learning to resolve poor data ‘quality’

Associated to the area of excellence Risk & Finance.


The use of machine learning comes up against ‘data scarcity’, i.e. poor data ‘quality’, which can only be exploited if it has been ‘labelled’ before and according to the analysis you want to carry out. 


The method developed will improve the performance of predictive models and reduce the costs (in terms of time and resources) of data labelling.


The pre-training of predictive models makes it possible to overcome the limits inherent in data quality. We are developing a method called pre-learning, aiming to:

  • Identify and collect all available data;
  • Generate and obtain a deep learning model synthesising all the information extracted;
  • Apply this new resource to meet business challenges (e.g. risk assessment, fraud detection).


Companies in all sectors.


Doctor Anass Akrim

Doctor Anass Akrim

Anass obtained his PhD in Artificial Intelligence and Applied Mathematics in predictive maintenance (aeronautics sector). He studied Applied Mathematics, Computer Science and Finance at Paris Dauphine University. He also has an engineering degree from the Ecole des Mines in Big Data and Data Science, combined with a double degree in Banking and Finance from the IAE in Saint-Etienne. Anass has worked in a variety of business sectors and on a range of A.I. applications: time series processing (stock price prediction, automatic trading), financial fraud detection, predictive maintenance (aeronautics sector).

  • Article académique : A. Akrim et al. “Self-Supervised Learning for Data Scarcity in a Fatigue Damage Prognostic Problem” Engineering Applications of Artificial Intelligence, 2023, doi :
  • Note : “L’intelligence artificielle dans le secteur financier, le défi de la « Data Scarcity » ”
  • Working paper : “L’IA pré-entraînée : une technique innovante pour une détection de fraudes plus performante” (à venir)


Share This