電気通信システムと管理

E-BABE-Fast Pricing Exotic products using machine Learning Techincs

Abstract

CHAOUACHI Wassim

Structured products are becoming more and more important in the world of investment banking,  and  more  and  more  investors  are  incorporating  this  type  of  asset  in  their portfolios.  There are various types of structured products, suitable for different investor profiles, including individual.

The objective of this article is to introduce new pricing methods other than Monte Carlo methods to speed up the computation time of some structured products called exotic products.  We will show how we reduce computation time from 371 days to 2.11 seconds keeping a very accurate precision. First  we  will  introduce  the  financial  products  we  price,  for  that  we  will  describe the  environment.  Second,  knowing  that  we  never  used  machine  learning  technics  to  price  products  at HSBC, one of the parts of the projects was a proof of concept on vanilla products to see if we can apply such techniques (Machine Learning) on more complex products such as Exotics. Third, We will introduce a new deep learning model for non linear interpolation to price Exotic products:  Autocallables for Mono-Underlying then on Multi-Underlyings .The last part of this project was the back-testing of our model on the last months .