Browse Prior Art Database

Capitalizing on Information Disparity in Local Financial Markets

IP.com Disclosure Number: IPCOM000248030D
Publication Date: 2016-Oct-19
Document File: 2 page(s) / 322K

Publishing Venue

The IP.com Prior Art Database

Abstract

Disclosed is a method that aims to close the gap between geographical variance in financial coverage. The method utilizes teleprompter feeds and/or closed captioning data fed through a Machine Learning and Natural Language Processing pipeline to extract relevant features for analysis.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 51% of the total text.

Page 01 of 2

Capitalizing on Information Disparity in Local Financial Markets

The disclosed method aims to help mitigate one of the primary inconsistencies in financial analysis: geographic coverage. As outlined by Engelberg and Parsons [1] in the past 5 years, there is a significant (causal) influence on capital markets that is region specific, e.g. local Minnesota retail investors may drive an increase in value for a local company by 50% overnight, but in North Carolina where coverage of this local Minnesota company is non-existent or minimal investors would miss out on this opportunity entirely.

Essentially, the local coverage of certain financial events limits the analysis of aggregate market outcomes.

However, to just "start covering" local markets presents myriad difficulties for financial firms. Among them are: power outages due to unforeseen weather events, delayed news cycles between time zones, and inconsistent reporting across outlets within the same local region but within different subregions.

This method aims to close the gap between geographical variance in financial coverage utilizing teleprompter feeds (and/or closed captioning data) to better predict the aggregate market. This algorithm seeks to do so in a way that mitigates the problems mentioned above.

The method takes teleprompter feed data from local news media, pipes it into a Natural Language Processing pipeline, and then reorganizes the cluster centers of the output around certain trends in the market. Underlying the method is an algorithm that adapts the k-means algorithm (Lloyd's algorithm) [2] such that k-initial is not randomly generated (standard approach to Lloyd's algorithm), but rather placed according to geographic distances between local subregions. The output is then re-clustered not according to dis...