Digitalization

• Banks and financial institutions have multiple streams where they transform data into information (mainly historical batch processes).
• ML/DL can find patterns in the data flow to predict the data quality of a single transaction. 
• In addition, ML/DL can advise how to solve the data issue (spanning from personalized emails up to automated corrections).

A comprehensive approach to risk management and tools for a modern automation of information processing of unstructured risk information.

• By identifying key value drivers and streamlining the planning process, FSI organizations can focus on key drivers to optimize its business model. 
• Using sensitivity analysis and approaches from the field of predictive analytics, ifb leverages its learnings to develop a key KPI’s library.

• Reference Framework for Stress testing in a multiperiod risk analysis 
• Enables a much better understanding and reflection of complex chain of effects in the capital planning process dependent on client needs
• Provides a comprehensive view on simulated forecasts for the economic capital using modern technology and modern methodologies (e.g. open source software, machine learning algorithms, Agent-based-modeling)
  Facilitates the ad-hoc capital planning
• Allows to compare the impact of different models on the results

• Machine Learning models can add significant improvements to the prediction power by using non-linear modeling.
• Using Tensorflow opens up for feature to improve the rating accuracy.

• The forecast of a bank's RWA (and in combination with capital development also the CET1 ratio1) has become a central management task.
• The (normative) capital planning requires projections of the RWA development over several years. A flash report becomes the first indication of RWA and thus the CET1 ratio for the current month.

• The use case also shows a practical application of graph databases and graph algorithms.
• The use case NFR shows how the different risk categories can be analyzed in a connected view, including risk driver analysis and compact management presentation.

• What are challenges and obstacles in fraud prevention in insurance claims management.
• The common algorithms (like autoencoder) and more complex anomaly detection algorithms can support to provide an optimal solution 

• Machine learning techniques are concerned with general pattern recognition or the construction of universal approximators of relations in the data in situations where no obvious a priori analytical solution exists. 
• As a new trend of machine learning, deep learning showed up at the end of last century and became one of the most efficient learning algorithms.
• Many of the popular Machine Learning & Deep Learning Frameworks are based on open-source.