Federated Intervention Prediction Across Privacy-Separated Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing techniques face difficulties in predicting the effect of intervention on a target when the status of intervention and the resulting outcome cannot be associated due to information exchange barriers between different organizations.
Innovation Solution
A federated learning approach is employed by multiple information processing devices to train prediction models independently, allowing them to predict intervention effects without sharing sensitive data, using features and outcomes from each device.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If federated learning is used to train prediction models independently on each device, then data privacy and security are maintained, but the ability to associate intervention status and results directly is lost
Solution Approach 1:
The patent introduces federated learning as an intermediary mechanism that enables collaborative training between organizations without direct data sharing. Each organization trains local prediction models using their own data (intervention status or results), and the models are aggregated through federated learning to achieve the same predictive capability as if all data were centralized, thus maintaining data privacy while preserving information association
Solution Approach 2:
The patent replaces the mechanical data exchange system (direct sharing of intervention status and results between organizations) with a model-based system. Instead of exchanging raw data, organizations exchange prediction models through federated learning, substituting the physical data transfer mechanism with a virtual model collaboration mechanism that achieves the same goal without compromising data security
2Measurement precision
If direct data exchange between organizations is implemented, then complete training data can be obtained, but data privacy and security are compromised
Solution Approach 1:
The patent extracts the essential predictive capability from the raw data itself. Instead of exchanging complete datasets (intervention status and results), each organization extracts and shares only the prediction model parameters through federated learning. This extraction approach maintains prediction accuracy while removing the privacy risk associated with sharing sensitive raw data
Solution Approach 2:
The patent changes the form of data exchange from raw observational data (intervention status and results) to model parameters (weights and biases of prediction models). This parameter transformation allows organizations to share predictive knowledge without exposing sensitive information, as the model parameters alone cannot reconstruct the original private data
Data Source
AI summary
A first information processing device includes a first acquisition unit acquiring a feature representing a target and a result obtained for the target, a first prediction unit predicting a status of intervention likely to have affected the result based on the feature, and a first prediction model training unit training a prediction model for predicting an effect of the intervention by federated learning performed by the first information processing device and a second information processing device based on the feature, the result, and the status of the intervention. The second information processing device includes a second acquisition unit acquiring the feature and the status of the intervention, a second prediction unit predicting the result based on the feature, and a second prediction model training unit training the prediction model by the federated learning based on the feature, the result, and the status of intervention.