Transfer Learning Model Selection for Temporary Data Shift Validation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional techniques fail to account for temporary and unintended changes in data distribution, leading to difficulties in model validation and factor analysis in machine learning models used in monitoring systems.
Innovation Solution
An information processing apparatus that stores model history, calculates evaluation values using new data, selects the most appropriate model for updating through transfer learning, and determines periods of accidental data changes, enabling easier model validation and factor analysis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional model updating techniques are used, then model updating can be performed, but model validation and factor analysis become difficult when data distribution changes temporarily and unintendedly
Solution Approach 1:
The patent applies preliminary action by detecting extraordinary periods (temporary and unintended data distribution changes) before they affect model validation. The system identifies these periods in advance and excludes them from the learning process, preventing them from causing validation difficulties or factor analysis errors.
Solution Approach 2:
The patent changes the parameter selection criteria for model learning by introducing a filter based on extraordinary period detection. Instead of using all available data, the system modifies the learning data selection to exclude data from detected extraordinary periods, thereby improving model validation reliability while adapting to temporary data distribution changes.
2Productivity
If all available data is used for model learning, then learning efficiency is high, but factors indicated by the model considerably change before and after extraordinary periods
Solution Approach 1:
The patent applies the extraction principle by removing data from extraordinary periods out of the total available data. The system detects these periods and extracts only the normal data for model learning, preventing the harmful influence of temporary data distribution changes while maintaining learning efficiency with high-quality data.
Solution Approach 2:
The patent converts the harmful effect of extraordinary periods into a benefit by using them as detection targets. The system identifies these periods through data distribution analysis and uses this detection to improve model stability, transforming what would be harmful noise into a useful filtering criterion for selecting better learning data.
3Ease of operation
If model updating is performed without considering extraordinary periods, then updating process is simple, but validation or factor analysis of the model is made difficult
Solution Approach 1:
The patent applies self-service by enabling the model updating system to automatically detect and handle extraordinary periods without manual intervention. The system autonomously identifies data distribution changes, determines extraordinary periods, and adjusts the learning data selection accordingly, maintaining operational simplicity while preserving validation information reliability.
Data Source
AI summary
An information processing apparatus according to one embodiment includes one or more hardware processors connected to a memory. The hardware processors functions to store, in the memory, history information including identification information of a model and a history of updating the model. The model receives input data including variables and outputs output data. The variables are each a variable for which a rate of influence on the output data is calculated. The model has been updated by using first input data. The hardware processors functions to select a target model to be updated by using second input data. The target model is selected from among models identified by their respective identification information. The hardware processors functions to update the target model by performing transfer learning in which updated parameters are estimated by using the second input data.