Development of a heterogeneous data repository for rating broker recommendations in capital markets

Institutions recommend financial products and instruments from different brokers to their customers. The institutions are required to ensure that the trades serve their customer interests. The existing rule-based system results in high false positives. The recommendations that are identified as positives are scrutinized by financial analysts, which is costly and inefficient. The project will enhance this rule-based automated system with a data repository to reduce the false positives. The system will not only capture the human intelligence, but also the actual rate of returns from the proposed trade. In addition to being useful to the financial analysts, the data repository is expected to be used for subsequent machine learning for reducing false positive rates. This AI based system will be a significant step forward that uses state-of-the-art clustering, natural language processing (NLP), sentiment analysis, fuzzy set theory to develop both unsupervised and supervised primary machine learning models. These primary models will rate recommendations based on their deviation from the market trends, social media sentiments, and tone of emails.

Faculty Supervisor:

Pawan Lingras

Student:

Partner:

Lexington Innovations Inc.

Discipline:

Computer science

Sector:

Finance and Insurance; Professional, scientific and technical services

University:

Saint Mary's University

Program: