User-Merchant Matching Algorithm with Privacy Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing matching algorithms fail to accurately match users with relevant merchants due to reliance on inferred user interests from web behavior, neglecting user-preferred data and changing interests over time, and lacking user control over data dissemination.
Innovation Solution
A method that receives user profile data including privacy settings and interest preference data, inputs this data into a machine learning algorithm to generate matches between users and merchants, and generates data sharing messages based on these matches, ensuring that only relevant and privacy-compliant data is shared with merchants.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing matching algorithms use inferred user interests from web behavior, then they can automatically suggest matches, but the matching accuracy decreases because user-preferred data is not considered
Solution Approach 1:
The patent segments user data into multiple types: explicit preference data (from direct user input), implicit behavior data (from web behavior), and contextual data. By segmenting data sources and processing them through separate pathways before integration, the system can accurately weigh explicit preferences more heavily while still incorporating behavioral insights, thereby improving matching accuracy without overwhelming complexity
Solution Approach 2:
The patent introduces a preference data intermediary layer that sits between raw user inputs and the matching algorithm. This intermediary processes and validates explicit preference data, reconciles it with inferred behavior data, and presents a unified user profile to the matching algorithm. This mediator structure improves accuracy by ensuring preference data is properly integrated while managing the complexity of handling multiple data sources
2Adaptability or versatility
If existing systems track user web behavior to infer interests, then they can continuously update user profiles, but they cannot efficiently detect changing user interests over time
Solution Approach 1:
The patent implements dynamic preference weighting where explicit user-stated preferences are given higher weight and lower time decay compared to inferred behavioral preferences. When users update their explicit preferences, the system dynamically adjusts the profile with immediate effect, allowing rapid adaptation to changing interests without waiting for behavioral patterns to emerge, thus reducing the time lag in detecting interest changes
Solution Approach 2:
The patent incorporates feedback loops where the system monitors both explicit preference updates and behavioral deviations. When explicit preferences change or when behavioral data consistently diverges from the current profile, the system triggers re-evaluation and rapid profile updates. This feedback mechanism enables efficient detection of interest changes by actively monitoring for signals of preference evolution rather than passively waiting
3Productivity
If existing algorithms suggest matches to all merchants, then they can maximize potential connections, but they create mismatches between suggested users and merchants
Solution Approach 1:
The patent applies local quality by tailoring the user profile and preference weights specifically for each matching context and merchant type. Rather than applying a single global matching rule, the system adjusts which preferences are most relevant based on the specific merchant category and user-merchant context. This localized approach improves match relevance by ensuring the right preferences are emphasized for each specific matching scenario while maintaining efficient processing through contextual shortcuts
4Adaptability or versatility
If systems distribute user data to all merchants, then they can enable broad data access, but they violate user privacy settings and control
Solution Approach 1:
The patent implements parameter-based data sharing where user profiles include explicit privacy parameters and consent flags for different data types and merchant categories. The system dynamically changes data sharing parameters based on user preferences, allowing flexible data access for some merchants while restricting access for others. This parameterized approach enables broad data sharing flexibility while preventing privacy violations by enforcing user-defined access controls at the data transmission level
Data Source
AI summary
Methods, systems, and computer program products are provided for improving matching algorithms. A method may include: receiving user privacy settings and interest preference data; storing the data in a database; inputting the interest preferences data into a machine learning algorithm to generate at least one match between the user and a merchant, wherein the merchant comprises a subset of a plurality of merchants; generating a data sharing message by: compiling first data associated with the user; filtering the first data based on interest preference data and/or privacy settings data to generate shareable data; and for each merchant associated with a subset of the data associated a merchant, generating a corresponding data sharing message containing a subset of interest preference data; and distributing the corresponding data sharing messages.