Machine learning architecture for risk modelling and analytics

Abstract

A computer-implemented method for forecasting the pricing of at least one financial instrument, the method comprising a processor and a memory, the method comprising the operations of: generating, by the processor, a user interface on a display, said user interface comprising a user-selectable pricing tab; wherein selecting the pricing tab causes the processor to at least: receive raw data in a plurality of disparate formats; scrub the raw data for anomalies and null values using a set of rules and generate a structured data set; and wherein selecting said pricing tab causes the processor to measure best-fit correlations with respect to a company's fundamental valuation and secondary market pricing for the company's at least one financial instrument across sector peers and market conditions and generate at least one financial instrument pricing output, in real-time.

1 . A computer-implemented method for forecasting the pricing of at least one financial instrument, the method comprising a processor and a memory, the method comprising the operations of:
 generating, by the processor, a user interface on a display, said user interface comprising a user-selectable pricing tab; wherein selecting the pricing tab causes the processor to at least:   receive raw data in a plurality of disparate formats;   scrub the raw data for anomalies and null values using a set of rules and generate a structured data set; and   wherein selecting said pricing tab causes the processor to measure best-fit correlations with respect to a company's fundamental valuation and secondary market pricing for the company's at least one financial instrument across sector peers and market conditions and generate at least one financial instrument pricing output, in real-time.   
     
     
         2 . The method of  claim 1 , wherein the processor extracts at least one feature vector set from the structured data set. 
     
     
         3 . The method of  claim 2 , wherein the structured data set is input into a machine learning architecture comprising a neural network, wherein the machine learning architecture generates a machine learning model using the at least one feature vector set. 
     
     
         4 . The method of  claim 3 , wherein the machine learning model is iteratively trained to calculate relative value pricing curves for at least one currency. 
     
     
         5 . The method of  claim 3 , wherein the machine learning model is iteratively trained to predict a price of the at least one financial instrument within a time frame. 
     
     
         6 . The method of  claim 1 , comprising the further steps of:
 displaying, by the processor, on the user interface a plurality of user-selectable criteria comprising time thresholds pertaining to at least one of historical data, contemporaneous data, a sector, a tenor, a bond rating, a model preference, a confidence level and a liquidity score;   displaying on the user interface the plurality of user-selectable criteria with selection indicators received by the processor to execute the instructions to forecast the pricing of the financial instrument.   
     
     
         7 . The method of  claim 6 , wherein the at least one financial instrument is a bond. 
     
     
         8 . The method of  claim 1 , wherein the user interface comprises a user-selectable matching tab, wherein selecting said matching tab causes the processor to employ algorithmic matching of target buyers with the at least one financial instrument, based on at least one of past buying patterns, portfolio manager preferences, rebalancing events and preferred industry sector, rating or tenor; and to generate a ranking score indicative of the target buyer's likelihood to purchase the at least one financial instrument. 
     
     
         9 . A computer-implemented method for forecasting the issuance of at least one financial instrument, the method comprising a processor and a memory, the method comprising the operations of:
 generating, by the processor, a user interface on a display, said user interface comprising a user-selectable issuance tab;   wherein selecting the issuance tab causes the processor to at least:   receive raw data in a plurality of disparate formats;   scrub the raw data for anomalies and null values using a set of rules and generate a structured data set; and   wherein selecting said issuance tab causes the processor to perform measurement of at least one financial instrument issuer's propensity to issue the at least one financial instrument, and to assign a propensity score which estimates the relative likelihood the issuer will issue the at least one financial instrument within a time frame; and   
       output an issuance recommendation for the issuer. 
     
     
         10 . The method of  claim 9 , wherein the processor extracts at least one feature vector set from the structured data set. 
     
     
         11 . The method of  claim 10 , wherein the structured data set is input into a machine learning architecture comprising a neural network, wherein the machine learning architecture generates a machine learning model using the at least one feature vector set. 
     
     
         12 . The method of  claim 11 , wherein the machine learning model is iteratively trained to calculate relative value pricing curves for at least one currency. 
     
     
         13 . The method of  claim 11 , wherein the machine learning model is iteratively trained to predict an issuance of the at least one financial instrument within the time frame. 
     
     
         14 . The method of  claim 13 , wherein the at least one financial instrument is a bond. 
     
     
         15 . The method of  claim 9 , comprising the further steps of:
 displaying, by the processor, on the user interface a plurality of user-selectable criteria comprising time thresholds pertaining to at least one of historical data, contemporaneous data, a sector, a tenor, a bond rating, a model preference, a confidence level and a liquidity score;   displaying on the user interface the plurality of user-selectable criteria with selection indicators received by the processor to execute the instructions to forecast the issuance of the financial instrument.   
     
     
         16 . The method of  claim 15 , wherein the user interface comprises a user-selectable matching tab, wherein selecting said matching tab causes the processor to employ algorithmic matching of target buyers with the at least one financial instrument, based on at least one of past buying patterns, portfolio manager preferences, rebalancing events and preferred industry sector, rating or tenor; and to generate a ranking score indicative of the target buyer's likelihood to purchase the at least one financial instrument. 
     
     
         17 . The method of  claim 11 , wherein the machine learning model is iteratively trained to match at least one target buyer with the issuer. 
     
     
         18 . A computer-implemented method for trading in securities, the method comprising a processor and a memory, the method comprising the operations of:
 monitoring, by the processor, current and historical secondary market trading levels of correlated securities;   receiving, by the processor, raw data comprising new issue pricing levels and secondary traded pricing levels from a plurality of dealers and capital markets data sources, wherein the raw data is in a plurality of disparate formats;   converting, by the processor, the plurality of disparate formats into a standardized format;   responsive to said monitoring, by the processor, predicting in real-time at least one of issue likelihood, new issue pricing levels and secondary traded pricing levels of a plurality of issuers of the securities and the new issue pricing levels and secondary traded pricing levels of the securities;   converting, by the processor, the new issue pricing levels and secondary traded pricing levels to equivalent levels in any one of a plurality of foreign currencies and any one of a plurality of interest rates.   
     
     
         19 . The method of  claim 18 , comprising a further step of:
 processing, by the processor, issue transactions following the matching step; and   continuously monitoring, by the processor, regulatory compliance and reporting mandates associated with the issue transactions and secondary market trading transactions.   
     
     
         20 . The method of  claim 18 , comprising a further step of:
 monitoring, by the processor, traditional and non-traditional buyer preferences, based on historical buying patterns of traditional and non-traditional buyers and aggregating, by the processor, buyer preferences from expressions of interest.   
     
     
         21 . The method of  claim 18 , comprising a further step of:
 generating, by the processor, regulatory or market monitoring reports related to deal and non-deal-related activities in the markets.   
     
     
         22 . The method of  claim 18 , comprising a further step of:
 determining, by the processor, current secondary market liquidity in real-time, and using aggregate data derived from public data and proprietary private user data, and based on a set of securities and a group of dealers.   
     
     
         23 . The method of  claim 18 , comprising a further step of:
 displaying, by the processor, at least one of historical trend analysis charts, historical deal analysis charts, and sector and peer comparison charts based on the aggregate data.   
     
     
         24 . The method of  claim 18 , comprising the further steps of:
 publishing, by the processor, the pricing levels using the standardized format to a plurality of users;   evaluating, by the processor, secondary market liquidity;   comparing, by the processor, covenant terms, and evaluating, by the processor, pricing profiles of the plurality of securities; and   wherein the step of predicting the at least one of issue pricing levels and secondary market security pricing levels of a plurality of issuers of the securities and secondary market pricing levels of the securities comprises the further steps of:   evaluating, by the processor, participation in primary and secondary capital markets by a plurality of buyers based on their established buying pattern analysis and simultaneously considering attractiveness of the current pricing levels.   
     
     
         25 . The method of  claim 18 , comprising further steps of:
 publishing, by the processor, pricing indications publicly or privately to other users among the plurality of users;   collecting and aggregating, by the processor, feedback data which is applied in a pricing and buying pattern analysis process; and   enabling, by the processor, the plurality of users to create a plurality of deals and populate the plurality of deals with a plurality of existing reverse inquiries, and gauging interest in the plurality of deals.   
     
     
         26 . The method of  claim 18 , comprising a further step of predicting, by the processor, at least one of an additional likely traditional and non-traditional buyer based on an established buying pattern including ranking based on their size, frequency and recency of purchases. 
     
     
         27 . The method of  claim 18 , comprising a further step of matching, by the processor, at least one target buyer to specific issuers based on the at least one of the predicted issue likelihood and the predicted new issue pricing levels and the predicted secondary traded pricing levels. 
     
     
         28 . A computer readable medium storing instructions executable by a processor to carry out the operations comprising:
 aggregating raw data from a plurality of data sources comprising contemporaneous trading data and fundamental data covering a series of time periods and one or more aspects of quantitative investing and market monitoring, said raw data in a plurality of disparate formats;   transforming said raw data in the plurality of disparate formats into a single standard format to generate structured data;   extracting at least one data element of value associated with at least one financial instrument from the structured data in accordance with one or more pre-programmed functions;   establishing a plurality of input nodes and an output node for a recurrent neural network model for each aspect of quantitative investing and market monitoring;   using the recurrent neural network model to build at least one model;   inputting the structured data into the recurrent neural network model using the plurality of input nodes;   training each of the recurrent neural network using said inputs until an error function associated with an output value that corresponds to an aspect of quantitative investing and market monitoring is minimized; and   using one or more weights from the trained recurrent neural network models to identify a set structured data by element of value and output that will be used as an element of value summary for use as an input to each of one or more predictive models;   normalizing each of the one or more sets of structured data by data element of value, refining the sets of structured data by the data element of value,   creating a summary of a refined transaction data set for each data element of value, and   using the data element of value summaries as inputs to a predictive model for each of the one or more aspects of quantitative investing and market monitoring where the aspects of quantitative investing and market monitoring comprising managing trading activities, managing risk, making portfolio funding allocations, predicting a time horizon for issuance of the at least one financial instrument; predicting an issuer of the at least one financial instrument within the predicted time horizon; predicting a price of the at least one financial instrument, matching a buyer with the at least one financial instrument, and combinations thereof, and   wherein the predictive models are useful for completing tasks comprising managing trading activities, managing risk, making portfolio funding allocations, predicting a time horizon for issuance of the at least one financial instrument; predicting an issuer of the at least one financial instrument within the predicted time horizon;   
       predicting a price of the at least one financial instrument, matching a buyer with the at least one financial instrument, and combinations thereof. 
     
     
         29 . The computer readable medium of  claim 28 , wherein the recurrent neural network comprises a bidirectional long short-term memory (BLSTM) neural network architecture. 
     
     
         30 . The computer readable medium of  claim 28 , comprising the further operations of:
 displaying, by the processor, on the user interface a plurality of user-selectable criteria comprising time thresholds pertaining to at least one of historical data, contemporaneous data, a sector, a tenor, a bond rating, a model preference, a confidence level and a liquidity score;   displaying on the user interface the plurality of user-selectable criteria with selection indicators received by the processor to execute the instructions to forecast the pricing of the financial instrument.   
     
     
         31 . The computer readable medium of  claim 28 , wherein the at least one financial instrument is a bond.