Probable Class Prediction via Target Probability Thresholds
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Machine learning models face significant challenges in classification due to misclassification errors, which can lead to erroneous predictions and undesirable consequences in real-world applications, particularly where clean separation of training data is not possible and zones of misclassification occur.
Innovation Solution
A method and system for predicting a set of probable classes for test data by determining membership probabilities based on user-input target probabilities, using an AI model that processes test data through feature extraction, distance calculations, and probability determination, ensuring a combined probability of selected classes exceeds the target threshold, thereby reducing misclassification risks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a machine learning model performs classification by predicting a single class for test data, then the classification process is simple and fast, but misclassification errors occur leading to erroneous predictions
Solution Approach 1:
The patent segments the classification output into multiple probable classes instead of a single prediction. The system divides the prediction space by identifying multiple candidate classes with their respective probability values, allowing the user to select from segmented options rather than accepting a single potentially erroneous classification.
Solution Approach 2:
The patent applies partial action by providing only the top-k probable classes rather than exhaustively listing all possible classes. This selective approach provides sufficient information for accurate classification while maintaining computational efficiency, avoiding the excessive action of analyzing every possible class equally.
2Reliability
If the system provides multiple probable classes with probabilities, then misclassification risks are reduced, but the output complexity and processing time increase
Solution Approach 1:
The system performs partial action by selecting only the top-k most probable classes for output. Instead of processing all possible classes equally, the algorithm identifies and presents only the necessary subset that achieves the user's target probability threshold, significantly reducing processing time while maintaining high prediction reliability.
Solution Approach 2:
The patent utilizes parameter changes by allowing the user to specify a target probability threshold. The system dynamically adjusts the number of classes returned based on this parameter, providing fewer classes when a higher threshold is set and more classes when a lower threshold is acceptable, thereby optimizing processing time according to reliability requirements.
3Reliability
If the system requires users to specify a target probability, then prediction reliability can be controlled, but the ease of operation decreases
Solution Approach 1:
The system implements dynamics by making the number of returned classes flexible and adaptive. Instead of a fixed output format, the system dynamically adjusts the prediction output based on user-specified reliability requirements, allowing the same interface to adapt to different user needs without requiring complex configuration.
Solution Approach 2:
The patent applies parameter changes by using a simple numeric target probability input that fundamentally changes the system's behavior. This single parameter controls the balance between reliability and output size, making the complex prediction system controllable through a simple user-friendly interface rather than complex settings.
Data Source
AI summary
A system and method for predicting a set of probable classes for test data is described. The method comprises retrieving, from a memory a training dataset, a plurality of classes, and a plurality of corresponding training feature vectors for each of the plurality of classes. The method comprises receiving an input indicative of a target probability required for the test data. The method comprises determining a set of membership probabilities for the test data that include a corresponding membership probability associated with each of the plurality of classes, the corresponding membership probability being indicative of a probability of the test data belonging to a corresponding class of the plurality of classes. The method comprises determining, based on the input and the set of membership probabilities, the set of probable classes, from the plurality of classes, for the test data.