Automated Rule Learning Framework for ML Prediction Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current machine learning and deep learning techniques face challenges in accurately classifying small datasets, often resulting in incorrect data classification and requiring time-consuming, manual rule development, which is not comprehensive and struggles with noisy text data.
Innovation Solution
An automated rule learning framework that generates rules using entropy values and model explainability techniques, allowing for the improvement of machine learning models by segmenting and clustering data, and incorporating semantic rule builders to enhance prediction performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If machine learning or deep learning techniques are used to classify small datasets, then automation is improved, but classification accuracy deteriorates
Solution Approach 1:
The patent segments the classification task into multiple components: initial ML/DL classification, entropy calculation for each segment, rule generation based on entropy values, and iterative refinement. This segmentation allows the system to handle small datasets by breaking down the complex classification problem into manageable steps that progressively improve accuracy.
Solution Approach 2:
The patent introduces entropy values and automatically generated rules as intermediary elements between the ML/DL model and the final classification result. These intermediaries bridge the gap by providing additional structure and guidance that helps the model achieve better classification accuracy on small datasets without sacrificing automation.
2Measurement precision
If manual rule development is used to improve classification accuracy, then measurement precision is improved, but productivity deteriorates
Solution Approach 1:
The system performs self-service by automatically generating classification rules based on entropy calculations from the data itself. Instead of requiring manual expert intervention, the system autonomously analyzes the data patterns, calculates entropy values, and generates appropriate rules, thereby maintaining high classification accuracy while dramatically improving productivity.
Solution Approach 2:
The patent replaces the manual mechanical process of rule development with an automated computational system. The entropy-based rule generation mechanism substitutes human experts' manual rule creation, achieving both high precision and improved efficiency through algorithmic automation.
3Productivity
If automated rule generation is implemented, then productivity is improved, but device complexity deteriorates
Solution Approach 1:
The system performs preliminary actions by pre-calculating entropy values for different data segments before generating rules. This preliminary analysis simplifies the subsequent rule generation process and makes the overall system more manageable despite the added automation capabilities.
Solution Approach 2:
The patent manages complexity by changing parameters systematically - using entropy values as a key parameter to control rule generation. By adjusting entropy thresholds and focusing on high-entropy segments, the system maintains manageable complexity while achieving automated rule development.
4Manufacturing precision
If clustering of centroid samples is performed on large datasets, then manufacturing precision is improved, but use of energy deteriorates
Solution Approach 1:
The patent extracts and processes only the most relevant portions of large datasets by performing clustering on centroid samples rather than the entire dataset. This extraction approach maintains high processing accuracy by focusing on representative samples while significantly reducing computational energy consumption.
Solution Approach 2:
The system applies partial action by performing clustering only on a subset of centroid samples rather than the complete dataset. This partial processing approach achieves sufficient precision for the application while avoiding the excessive energy consumption that would result from processing all data points.
Data Source
AI summary
The disclosure relates to a method for receiving labelled data from data sources; generating a plurality of segments from the labelled data using tokenizers, wherein each of the segments is associated with one or more classes, and wherein the plurality of segments comprises a sub-plurality of unique segments; calculating an entropy value for each of the sub-plurality of unique segments; for each class, generating one or more rules based on each of the segments associated with the class and the entropy value of each such segment; and combining the generated one or more rules with a model to improve prediction performance of the model. The labelled data is input data for a classifier model to form predictions and false positives are extracted from the predictions. The model explainability techniques and the false positives are used to create additional rule.