Inverse Signal Classification for Financial Instruments

TL;DR

This paper introduces innovative machine learning techniques for classifying financial time-series data, capable of handling diverse lengths and types, and demonstrates their application in identifying inverse behaviors among thousands of financial instruments.

Contribution

The paper presents novel signal composition and self-labeling methods that improve classification of complex financial time-series data, with implementation details for a financial search engine system.

Findings

Effective classification of diverse financial time-series

Identification of inverse behaviors among instruments

Scalability to large datasets of 7,881 instruments

Abstract

The paper presents new machine learning methods: signal composition, which classifies time-series regardless of length, type, and quantity; and self-labeling, a supervised-learning enhancement. The paper describes further the implementation of the methods on a financial search engine system using a collection of 7,881 financial instruments traded during 2011 to identify inverse behavior among the time-series.