Black Swan Algorithm Implementation

Implemented black swan algorithm for a Japanese financial firm to figure out potentially fraudulent companies

Project Synopsis

This Financial Services startup firm wanted a solution for the Japanese Financial Services market; essentially a tool for Financial analysts, researchers and traders to keep track of suspicious corporate entities, and their networks.

The platform had to maintain all the information of such suspicious entities and their networks by constantly sourcing data from financial and market data providers, financial web-sites, social networking sites and blogs.

Our Solution

We worked closely with the client's team and an iteratively build the platform based on a paper published by Stanford students.

This solution had following highlights:

  • Built web based platform for maintaining entity information.
  • Financial data was sourced through Tokyo Stock Exchange’s TDNet and EDINet data feeds. This data is published by TSE as XBRL (eXtensible Business Reporting Language) files.
  • Unstructured data about an entity was sourced from corporate news sites, blog sites and social networking sites
  • The corruption/risk scores were calculated for all the entities in the database using probabilistic belief propagation algorithms.

Project Highlights

  • The product was delivered meeting full functional, performance, quality and timeline requirements of the client.
  • The proprietary risk algorithm gave an accurate view of the extent of corruption of each entity in the database, and it’s influence on other related entities.
  • Created a Ravis based network visualization of companies and their relations

About The Project

Implementation of an algorithm based on Stanford white paper about propagating relationship scores

Technologies Used

Java Based Backend
Natural Language Processing (NLP)
Ravis Visualization Framework
Custom Rules Engine

Client Details