Expert Rule Prioritization for Faster Transaction Categorization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The challenge of efficiently categorizing large volumes of transaction data with varying characteristics and complexities in automated transaction analysis systems, due to the volume and variability of electronic transaction data, is complex and inefficient.
Innovation Solution
An expert system model that employs multipattern substring matching, rule filtering, and rule prioritization techniques to reduce the number of rules required for categorizing transactions by automatically filtering and prioritizing rules, using closed-loop iterative statistical rule analysis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional rule-based categorization is used for transaction data, then categorization coverage is comprehensive, but system complexity and processing time increase significantly
Solution Approach 1:
The patent extracts and removes redundant rules from the rule set through automated analysis. The system identifies rules that are subsets of other rules or rules that never match any transactions, then removes these redundant rules to reduce system complexity while preserving categorization functionality.
Solution Approach 2:
The patent performs preliminary rule prioritization and filtering before the actual categorization process. By pre-ordering rules based on their importance and pre-identifying redundant rules, the system reduces the effective rule set size before processing transactions, thereby improving processing efficiency.
2Measurement precision
If more rules are added to cover all transaction variations, then categorization accuracy improves, but processing time increases
Solution Approach 1:
The patent extracts and removes redundant rules that do not contribute to categorization accuracy. By identifying rules that are always superseded by other rules or rules that never match transactions, the system eliminates unnecessary processing steps while maintaining accurate categorization.
Solution Approach 2:
The patent performs preliminary rule prioritization to determine the most important rules before processing. By pre-ordering rules based on their categorical importance and matching frequency, the system ensures that high-value rules are evaluated first, reducing average processing time while maintaining accuracy.
3Productivity
If rule set is reduced for faster processing, then processing efficiency improves, but categorization coverage may be lost
Solution Approach 1:
The patent carefully extracts only truly redundant rules while preserving all necessary categorization rules. The automated analysis ensures that rules removed are either superseded by other rules or never match transactions, thereby maintaining complete categorization coverage while reducing rule set size.
Solution Approach 2:
The patent performs preliminary analysis to identify and preserve high-priority rules that provide the most valuable categorization coverage. By pre-determining which rules are essential for maintaining coverage, the system can safely reduce the rule set while ensuring that critical categorization functionality is preserved.
Data Source
AI summary
A computing system receives rules and transaction data from a database. The system creates a potential rule list and stores the rules in the list on the database. The system creates an ordered rule list, a rule-to-transaction lookup table, and a transaction-to-rule lookup table. The system applies each rule to each transaction and identifies each transaction for which each rule is true. The system adds each rule and the identified transactions to the lookup tables and generates a priority-ordered list of the rules iteratively until all rules have been removed from the potential rule list. In each iteration, the system calculates an accuracy of each rule based on all transactions for which the rule is true, appends the highest accuracy rules to the ordered rule list, removes the rules from the potential rule list, and removes the transactions from all entries in the rule-to-transaction lookup table.