Query Classification Using SVM Probability Normalization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing query classification systems face challenges when dealing with multiple classes, as they often result in ties due to equal vote distribution, leading to ambiguity in determining the top class.
Innovation Solution
A server and method utilizing multiple support vector machine models to generate votes and calculate probabilities for each class, with the top class determined by the greatest number of votes and normalized support vector machine scores, employing a sigmoid function and binary support vector machine calculations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple support vector machine models are applied to generate votes for each class, then classification accuracy is improved, but tie situations occur when equal number of votes are received by multiple classes
Solution Approach 1:
The patent transforms the discrete vote count parameter into a continuous probability parameter by applying sigmoid function normalization to support vector machine scores. This parameter transformation allows for fine-grained differentiation between classes that would otherwise receive equal integer votes, thereby resolving ties while maintaining classification accuracy.
Solution Approach 2:
The patent introduces probability as an intermediary parameter between the discrete vote counts and the final classification decision. By calculating probabilities based on normalized support vector machine scores and combining them with vote counts, the system creates a nuanced ranking mechanism that prevents ties and provides confidence measurements for classification results.
2Measurement precision
If support vector machine scores are normalized using sigmoid function, then probability calculation is improved, but computational complexity increases
Solution Approach 1:
The sigmoid function is applied to transform raw support vector machine scores into normalized probability values between 0 and 1. This parameter transformation improves probability calculation precision by ensuring values are bounded and comparable, while the sigmoid function's mathematical properties allow for efficient computation using standard neural network libraries.
Data Source
AI summary
A server, method, and non-transitory computer readable medium for classifying a query into one of a plurality of classes are provided. The server includes a network interface, a memory storage unit and a processor. The method involves receiving a query applying a plurality of support vector machine models, calculating a probability, and determining a top class. The non-transitory computer readable medium is encoded with codes to direct a processor to carry out the method.