Rule Update Program for Machine Learning Model Interpretability
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Machine learning models based on rule sets face challenges in providing satisfactory reasonableness due to the inclusion of rules that are unlikely to be reasonable, making it difficult to enhance interpretability and reasonableness, especially when the number of rules increases exponentially, complicating calculation and reducing the likelihood of listing more reasonable models.
Innovation Solution
A rule update program and method that allows for re-mining of training data to replace or update initial rules with new alternatives based on user specifications, focusing on samples corresponding to initially specified rules, thereby improving the reasonableness and interpretability of the machine learning model.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If exhaustive mining is executed to extract a rule set from training data, then the completeness of rule extraction is improved, but the computational complexity and time required increase exponentially
Solution Approach 1:
The patent segments the exhaustive mining process into multiple stages: initial mining to generate candidate rules, user specification to identify unreasonable rules, sample detection to find corresponding training data, and re-mining to generate alternative rules. This segmentation reduces computational complexity by avoiding single-stage exhaustive processing while maintaining rule extraction completeness.
Solution Approach 2:
The patent performs preliminary user specification of unreasonable rules before executing re-mining operations. This preliminary action allows the system to focus computational resources only on generating alternatives for identified problematic rules, rather than performing exhaustive mining on all possible rules, thereby reducing overall computational complexity.
2Reliability
If the number of rules in the rule set increases, then the coverage and explanatory power of the machine learning model are improved, but the interpretability and reasonableness deteriorate due to inclusion of unreasonable rules
Solution Approach 1:
The patent implements a feedback mechanism where user specifications about unreasonable rules are fed back into the system to guide re-mining operations. This feedback loop allows the system to iteratively improve rule quality by replacing unreasonable rules with alternatives that better satisfy user criteria, thereby maintaining interpretability while preserving coverage.
Solution Approach 2:
The patent changes the parameter set used for mining by executing re-mining with modified conditions based on user specifications. This parameter change allows generation of alternative rules that differ from original unreasonable rules, improving reasonableness while maintaining the necessary coverage for model explanatory power.
3Reliability
If re-mining is executed for all sample data, then the thoroughness of rule alternative generation is improved, but the computational burden increases significantly
Solution Approach 1:
The patent extracts and processes only the subset of sample data that corresponds to unreasonable rules identified through user specification. This extraction principle allows the system to perform re-mining on a reduced data subset rather than all training data, significantly reducing computational burden while maintaining thoroughness for the critical problematic rules.
Data Source
AI summary
A non-transitory computer-readable recording medium storing a rule update program for causing a computer to execute a process including: accepting user specification for at least a part of rules included in a rule set generated as a result of first mining that uses training data; detecting, from the training data, sample data that corresponds to the rules for which the user specification has been accepted; and acquiring a new rule by executing second mining by using the training data limited to the sample data that corresponds to the rules for which the user specification has been accepted.