Dynamic Data Classification Model Updating via User Curation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data classification methods based on static models and algorithms often produce high false positives and false negatives due to unpredictable, dirty, or ambiguous data, and require extensive user training with large datasets, which is not always feasible.
Innovation Solution
A system that allows users to create and associate classifications with data sets, compares classification models between data sets, and updates these models based on user curation actions, enabling more accurate data classification with limited user interaction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If machine learning models are used to improve data classification accuracy, then classification precision is improved, but the requirement for extensive user training and large amounts of training data increases
Solution Approach 1:
The system performs preliminary actions by automatically generating initial classification models using data properties and patterns before user interaction. This allows the system to have a head start in classification accuracy without requiring users to provide extensive training data upfront. The preliminary classification models are then refined through user feedback rather than requiring extensive pre-training.
Solution Approach 2:
The system serves itself by automatically learning from user curation actions and improving its own classification models without requiring external retraining. The system uses user feedback to iteratively update its classification models, making the system self-improving rather than requiring continuous external training input.
2Ease of operation
If static models and algorithms are used for data classification, then the system is simple to operate, but false positives and false negatives increase
Solution Approach 1:
The system transitions from static classification models to dynamic models that automatically adapt and update based on user feedback. The classification models evolve over time by learning from user curation actions, allowing the system to maintain simplicity of operation while continuously improving accuracy through automated model updates rather than requiring complex user configuration.
Solution Approach 2:
The system implements a feedback mechanism where user curation actions (corrections, validations, rejections) are automatically used to update and improve classification models. This feedback loop allows the system to learn from user interactions and improve classification accuracy without requiring users to manually configure complex parameters, maintaining ease of operation while reducing false positives and negatives.
3Measurement precision
If extensive user training with large amounts of training data is performed, then classification model accuracy is improved, but the time and resources required increase significantly
Solution Approach 1:
The system performs self-training by automatically learning from user curation actions in production environments. Instead of requiring extensive upfront training time, the system continuously improves its models using real user feedback, distributing the training process over time and eliminating the need for concentrated training sessions that consume significant user time and resources.
Solution Approach 2:
The system uses real-time feedback from user interactions to iteratively improve classification accuracy. Rather than requiring large batches of pre-collected training data and extensive training time, the system learns continuously from each user interaction, making the accuracy improvement process more efficient and less time-consuming while maintaining or improving model performance.
Data Source
AI summary
In some examples, a system receives a user input to create a classification to use for a first data set. In addition, the system may receive a user input to associate the classification with the first data set as reference data. The system may determine a classification association with the first data set based in part on comparing a first classification model corresponding to the reference data with a second classification model of a second data set, the second classification model being determined based at least in part on a plurality of data properties of the second data set. Further, the system may determine a user curation result with respect to the classification association with the first data set, and may update the classification model for the first data set based at least in part on the user curation result.