Dual Classification System with Dynamic Rule Updates
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing classification methods are imprecise and sluggish in adapting to new or unknown data, requiring large amounts of high-quality training data and being inflexible, especially in dynamic environments like spam filtering and cancer diagnostics.
Innovation Solution
A method that combines two complementary classification methods, such as outlier detection and rule-based methods, for improved accuracy and adaptation, where each method updates its criteria based on feedback to minimize false positives and negatives, allowing for continuous improvement and flexibility.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If machine learning methods are used for classification, then classification accuracy can be improved, but large amounts of high-quality training data are required and the system is slow to adapt to new data
Solution Approach 1:
The patent segments the classification task into two distinct parts: an online classification component that handles new data rapidly using simple rules, and an offline training component that periodically retrains the model using accumulated data. This segmentation allows the system to maintain high adaptability while gradually improving accuracy through periodic retraining, rather than requiring large training datasets upfront.
Solution Approach 2:
The patent implements a dynamic classification system where the decision boundary and classification rules are continuously updated based on new incoming data. The system transitions from static pre-trained models to dynamic adaptive rules that evolve over time, enabling rapid adaptation to new data patterns without requiring complete retraining.
2Adaptability or versatility
If expert systems are used for classification, then flexibility and adaptability are improved, but implementation complexity increases
Solution Approach 1:
The patent implements a self-service system where the classification rules automatically generate and update themselves based on incoming data patterns, without requiring complex expert system configurations. The system serves itself by automatically learning decision boundaries and classification rules from data, reducing implementation complexity while maintaining flexibility.
Solution Approach 2:
The patent changes the parameters of the classification system dynamically based on incoming data, rather than relying on fixed expert knowledge rules. By adjusting classification thresholds and decision boundaries as parameters based on observed data patterns, the system achieves flexibility with simpler implementation compared to traditional expert systems.
3Device complexity
If existing classification methods are used, then implementation is simpler, but classification accuracy is reduced with more misclassifications
Solution Approach 1:
The patent incorporates feedback mechanisms where classification results are continuously monitored and used to refine and update classification rules. Misclassifications trigger automatic rule adjustments and retraining processes, enabling the system to improve accuracy over time while maintaining relatively simple implementation through automated feedback loops rather than complex algorithms.
Data Source
Figure 1
Figure 2a~2c
Figure 3a~3c
AI summary
Embodiments provide a method for classifying information into a first class or a second class. The method comprises a step of applying a first classification technique to the information in order to assign the information to the first class if the information satisfies classification criteria for the first class and in order to assign the information to the second class if the information does not satisfy the classification criteria for the first class. The method also comprises a step of applying a second classification technique to the information in order to assign the information to the second class if the information satisfies classification criteria for the second class and in order to assign the information to the first class if the information does not satisfy the classification criteria for the second class. The method also comprises a step of updating the classification criteria of at least one of the two classification techniques in the event of differing assignments of the information by the two classification techniques or in the event that a predefined number of differing assignments of information by the two classification techniques is reached. The first class and the second class are different.