Training Data Clustering for Classification Model Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Data classification models face performance issues due to the presence of mislabeled and irrelevant training data, especially in imbalanced datasets, which can lead to poor classification accuracy and increased computational resources spent on unauthorized transactions.
Innovation Solution
A model generation system that selectively removes mislabeled and irrelevant training data by clustering and analyzing the ratio of classifications within the data, using techniques like Bayesian Gaussian Mixture Models and variational autoencoders to improve the quality of the training data, thereby enhancing the performance of the classification model.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If training data is used as-is without filtering, then the training process is simple and fast, but the classification model performance deteriorates due to mislabeled and irrelevant data
Solution Approach 1:
The patent applies preliminary action by performing data filtering and quality assessment before the classification model training process. The system analyzes training data to identify and remove mislabeled or irrelevant samples prior to model training, ensuring that only high-quality data is used for training. This preliminary data preparation step improves classification accuracy while avoiding the need to retrain models with contaminated data.
2Loss of energy
If all training data is processed equally, then the processing method is simple, but computational resources are wasted on noisy and mislabeled data
Solution Approach 1:
The patent applies the extraction principle by identifying and removing harmful elements (mislabeled and irrelevant training data) from the training dataset. The system extracts quality metrics from training samples and selectively removes those that do not meet quality thresholds. This extraction of low-quality data prevents waste of computational resources on processing noisy data during model training, thereby improving energy efficiency.
3Manufacturing precision
If training data quality is improved by removing noisy data, then classification model performance improves, but the data preparation process becomes more complex and time-consuming
Solution Approach 1:
The patent applies mechanics substitution by replacing manual or rule-based data quality assessment with automated machine learning techniques. The system uses supervised learning models to predict data quality metrics and automatically identify mislabeled or irrelevant training samples. This substitution of automated intelligent systems for manual processes improves training data quality while minimizing the time and effort required for data preparation.
Data Source
AI summary
Methods and systems for training a computer-based classification model for classifying data are presented. The computer-based classification model is configured to classify data into one of a plurality of classifications. An initial training data set for training the classification model is obtained. In some embodiments, the training data within the initial training data set is grouped into multiple clusters, and training data within one or more clusters having corresponding ratio between a first classification and a second classification below a threshold ratio is removed from the initial training data set to generate the modified training data set. The modified training data set, instead of the initial training data set, is used to train the classification model.