Visualizing Financial Transactions via Tree Hierarchy Bundling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current systems for detecting credit card fraud and identifying anomalies in financial transactions are inadequate, as they rely on traditional fraud modeling that does not learn trends, leading to delayed detection and increased financial exposure, and manual data entry errors often result in unbalanced books and incorrect financial evaluations.
Innovation Solution
A data analysis system that generates a visual display of financial transactions using a tree structure hierarchy, representing transactions as lines connecting sources and destinations with weighted averages of coordinates, allowing for the bundling and coloring of transactions to highlight anomalies and errors, facilitating rapid detection of fraudulent patterns and data entry mistakes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional fraud modeling algorithms are used to detect fraudulent transactions, then the system can process large volumes of electronic transactions, but the detection speed is slow and financial exposure increases
Solution Approach 1:
The patent replaces traditional rule-based fraud modeling algorithms with a neural network-based machine learning system. The neural network learns patterns from historical transaction data and automatically detects fraudulent transactions, eliminating the need for manual rule creation and updating. This substitution enables faster detection while maintaining or improving accuracy, as the neural network can process transactions in real-time and adapt to new fraud patterns automatically.
Solution Approach 2:
The neural network system is self-learning and automatically improves its detection capabilities over time. It continuously processes transaction data, learns from confirmed fraud cases, and updates its internal models without requiring external intervention. This self-service mechanism allows the system to adapt to evolving fraud patterns autonomously, reducing detection time while maintaining high reliability.
2Ease of manufacture
If manual data entry processes are used for financial transactions, then data can be entered into back-office systems, but errors result in unbalanced books and incorrect financial evaluations
Solution Approach 1:
The patent replaces manual data entry processes with automated optical character recognition (OCR) and image processing technology. The system captures images of documents, automatically extracts transaction data using OCR, and validates the extracted information against predefined rules. This substitution eliminates human errors in data entry while maintaining ease of process execution, achieving both high speed and high accuracy in data capture.
Solution Approach 2:
The patent introduces an intermediary layer between document capture and data storage, consisting of OCR processing and validation rules. This intermediary automatically extracts data from images, verifies its accuracy through checksum validation and cross-referencing, and only allows corrected data to be stored in the back-office systems. This intermediary mechanism prevents erroneous data from entering the financial records while maintaining workflow efficiency.
Data Source
AI summary
A data analysis system (1) for displaying data facilitating visual analysis of financial transaction is disclosed. The system includes a transactions database (3) operable to store transaction records (5) defining financial transactions and a processing module (11) operable to determine a hierarchy having a tree structure wherein leaf nodes in the lowest level of the hierarchy correspond to sources and destinations associated with financial transactions represented by transaction records 5) stored in the transactions database (5). The processing module (11) then causes representations of the financial transactions to be displayed on a display screen (13) by determining for each transaction a first set of control co-ordinates comprising co-ordinates associated with elements in a path in the tree structure connecting the source and destination associated with a financial transaction via the closest common parent in the hierarchy common to the source and destination; determining for each transaction a second set of control co-ordinates for drawing a straight line between co-ordinates associated with the source and destination associated with the financial transaction; calculating as a set of control co-ordinates for representing a transaction weighted averages of corresponding co-ordinates in the first and second set, weighted by a bundling factor; and representing each of the financial transactions as a line drawn utilizing the calculated control co-ordinates for each transaction.