Financial Portfolio Reconciliation using Deconstructed Deep Learning
Citco provides financial products and services to hedge funds, private equity and real estate firms, investors, institutional banks, Global 1000 companies, and high net worth individuals in the Netherlands and internationally. The previous research was focused on optimizing operations by automating trade resolution and reducing risk using machine learning. Given its success in matching individual transactions, the proposed project plans to extend the research to overall portfolio management for identifying discrepancies in the portfolio at the end of the trading day. Outlier detection algorithms will be proposed to improve the accuracy of existing mismatch detection procedures. Unsupervised and supervised machine learning will be used to further improve this capability. The research will also look at improving data visualization and user interface for efficient monitoring of trade mismatch. Further, we will explore use of text mining for automated resolution of trades based on e-mails exchanged between those involved in a trade. The endeavor will significantly improve the data analytics capabilities of Citco and will potentially result in an increased operational footprint in Atlantic Canada.