Inverse Signal Classification for Financial Instruments

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.