Automated Data Validation System for Financial Prospectus Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Financial institutions face a time-consuming and error-prone manual process for validating data changes from third-party vendors, which are critical for investment decisions, requiring significant expertise and occurring asynchronously with portfolio construction.
Innovation Solution
A data validation system that automatically processes input data from third-party vendors using a multiclass classification model trained with natural language processing techniques to predict classification labels and generate confidence levels, enabling systematic and efficient validation of data changes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If manual data validation process is used, then data accuracy can be verified through expert judgement, but the process becomes time-consuming and error-prone
Solution Approach 1:
The patent replaces the manual mechanical validation process with an automated machine learning system. The multiclass classification model automatically analyzes prospectus data and validates financial entity information, substituting human expert judgement with algorithmic processing. This resolves the contradiction by providing both speed (automated processing) and reliability (consistent algorithmic application) without the time costs and human errors associated with manual validation.
Solution Approach 2:
The system enables self-service validation where the machine learning model independently processes and validates data without requiring human intervention for each validation task. The model trains on historical data and then autonomously validates new prospectus information, freeing experts from routine validation work while maintaining high accuracy through the model's learned patterns.
2Reliability
If manual validation by data analysts is performed, then qualitative judgement can be made on data changes, but significant expertise and training are required
Solution Approach 1:
The patent substitutes the complex human expert system with a machine learning model that encapsulates validation expertise in algorithmic form. The model learns from training data what constitutes valid versus invalid data changes, replacing the need for analysts to possess specialized knowledge. This resolves the contradiction by maintaining reliable validation judgement while eliminating the complexity of human expertise requirements.
Solution Approach 2:
The system creates a computational copy of expert validation knowledge through the machine learning model. By training on historical validation decisions and outcomes, the model captures and reproduces expert judgement patterns in a form that can be applied consistently without requiring the original human expertise. This allows the organization to retain validation capability even as individual experts leave or become unavailable.
3Ease of operation
If asynchronous validation process is used, then data analysts can review changes independently, but they do not know if instruments can be part of portfolio construction
Solution Approach 1:
The patent merges the previously separate validation process with the portfolio construction workflow. The machine learning model validates prospectus data changes and immediately integrates results with portfolio construction information, allowing analysts to see both validation status and portfolio impact in a unified system. This resolves the contradiction by maintaining operational simplicity while preventing information loss about portfolio context.
Solution Approach 2:
The system implements feedback loops where validation results are immediately communicated back to the portfolio construction process. The machine learning model not only validates data changes but also provides information about whether validated instruments can be included in portfolio construction, creating a continuous feedback cycle that keeps analysts informed of both validation status and portfolio implications without adding operational complexity.
Data Source
AI summary
A computer-implemented method is provided for validating input data from a third-party vendor. The method includes receiving, by a computing device, a plurality of prospectuses from a plurality of third-party entities and generating, by the computing device, a trained machine learning model using the plurality of prospectuses. The method also includes applying, by the computing device, the trained machine learning model on the input data to predict a classification label for the input data and generate a confidence level for the prediction.