Anomaly Detection Model Retraining for Data Trend Drift

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing anomaly detection algorithms using trained models face issues with over-detection and non-detection when data trends change, leading to repeated incorrect determinations.

Innovation Solution

A model training apparatus that includes a storage unit, detector, and retraining unit to store data and model parameters, determine anomalies, and retrain the model using corrected labels and trends to prevent repeated incorrect detections.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Speed

If a trained model is continuously used for anomaly detection, then detection speed is maintained, but detection accuracy deteriorates due to data trend changes

Engineering Contradiction:

Improvedetection speedVSAvoiddetection accuracy

Core Design Contradiction:

SpeedVSMeasurement precision

Solution Approach 1:

The system implements periodic retraining of the anomaly detection model at predetermined intervals. The retraining unit periodically updates the model using accumulated detection results and test data, ensuring the model adapts to changing data trends while maintaining continuous detection capability between retraining cycles.

Inventive Principle:

Principle #19Periodic action

Solution Approach 2:

The system uses detection results from the detector as feedback for model retraining. The retraining unit accumulates detection results and test data, then uses this feedback information to retrain and update the model parameters, creating a closed-loop system that continuously improves detection accuracy based on actual performance.

Inventive Principle:

Principle #23Feedback

2Measurement precision

If the model is retrained frequently to improve accuracy, then detection accuracy improves, but computational resource consumption increases

Engineering Contradiction:

Improvedetection accuracyVSAvoidcomputational resource consumption

Core Design Contradiction:

Measurement precisionVSUse of energy by moving object

Solution Approach 1:

Instead of continuous retraining, the system performs retraining periodically at predetermined intervals or when specific conditions are met. This reduces computational resource consumption compared to continuous retraining while still maintaining detection accuracy through periodic model updates.

Inventive Principle:

Principle #19Periodic action

Solution Approach 2:

The system uses accumulated detection results and test data over multiple periods for retraining, rather than processing all possible data continuously. This partial action approach achieves good detection accuracy while reducing the computational burden of processing complete datasets.

Inventive Principle:

Principle #16Partial or excessive action

Data Source

PatentUS12608617B2Model training apparatus, model training method, and program for retraining anomaly detection model

Publication Date: 2026.04.21 NT T INC

AI summary

An apparatus for training a model includes a storage unit configured to store a parameter of the model trained by using a training data set, and the training data set, a detector configured to use the model to determine whether an anomaly is present in a test data set and store a determined result and the test data set in the storage unit, and a retraining unit configured to retrain the model by using the determined result, the test data set, and the training data set.