Abnormal Pattern detection

Report Abstract:

In order to improve both the quality and the process of due diligence in the securitisation department, we set out a methodology to detect abnormal patterns. We start our analyse from raw data and check the asset quality before an investment.Our study applies both to the securitisation process for third-party and to investment for own account.

There are several ways to detect fraud. We will focus on different drivers for each steps of the securitisation process. Especially, the structuring and the business management of the deal. We set up fraud pattern with real fraud cases from the newspapers. From this point, we implement different test algorithms. Our results point out the high level of untapped or hardly untapped information on the company that appears in a dynamic analysis.

Research on insider Trading detection

You can find the final report in French here.

Speech recognition by neural network methods

All the details of this project are available here in French.