Federated Learning for Confidential Customer Profile Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Service providers face challenges in customizing their services for customers due to the confidentiality of customer data, which prevents the sharing of sensitive information between data holders and service providers.
Innovation Solution
A machine learning model is trained to select sales agents based on customer profiles that include both confidential and non-confidential information, predicting which agents have a high likelihood of making successful transactions with customers. This model compares customer profiles with sales agent profiles, considering attributes and previous transactions, to recommend the most suitable sales agents without disclosing confidential information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If customer data is shared with service providers for service customization, then service personalization improves, but data confidentiality and privacy protection deteriorate
Solution Approach 1:
The patent introduces federated learning as an intermediary mechanism that enables service providers to train AI models on customer data without directly accessing or storing the actual data. The model parameters and gradients are shared between the bank and service provider, allowing collaborative model training while keeping customer data confined within the bank's secure environment. This resolves the contradiction by enabling service customization through model collaboration without compromising data confidentiality.
Solution Approach 2:
The patent creates a virtual copy of the customer data environment through federated learning setups, where service providers work with local model copies rather than actual customer data. The service provider's model is trained on copies of the data distribution patterns without accessing the real sensitive information, enabling service personalization while maintaining data privacy through this virtual replication approach.
2Reliability
If customer data is not shared, then data security is maintained, but service providers cannot customize services effectively
Solution Approach 1:
Federated learning acts as an intermediary that decouples data security from service personalization. The bank's AI model serves as the intermediary asset that encapsulates customer data insights, allowing service providers to leverage these insights for personalization without direct data access. This enables service customization while maintaining data security through the model intermediary layer.
Solution Approach 2:
The patent segments the AI model training process into separate components: the bank holds and trains on customer data locally, while the service provider contributes to model improvement without accessing raw data. This segmentation of the training workflow allows both parties to contribute to service personalization while maintaining their respective data security boundaries.
3Measurement precision
If AI models are trained on confidential customer data, then prediction accuracy improves, but the risk of data exposure increases
Solution Approach 1:
The federated learning framework introduces model parameters and encryption protocols as intermediaries between the customer data and the training process. The bank's AI model processes customer data locally to generate updated model parameters, which are then securely transmitted to the service provider. This intermediary mechanism enables high prediction accuracy through continuous model training while minimizing data exposure risk through encrypted parameter exchange rather than raw data sharing.
Data Source
AI summary
Aspects described herein may allow managing profiles using machine learning models. A computing device may train a case-based reasoning (CBR) machine learning model using customer data, sales agent data, and completed transaction data, to predict a likelihood of a successful transaction between a customer associated with a customer profile, and a sales agent associated with a sales agent profile. After receiving a first customer profile, the computing device may determine, based on the confidential information and by inputting the first customer profile and the plurality of sales agent profile into the CBR machine learning model, a first sales agent, of a plurality of sales agents, that has a high likelihood of making a successful transaction with the first customer. The computing device may send an excerpt of the first customer profile omitting the confidential information.