Interactive ML Framework for Excluding Irrelevant User Activity Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current recommendation systems fail to address uncorrelated user data, leading to decreased accuracy and lack of user control over their data usage, despite user feedback mechanisms.
Innovation Solution
A system that allows users to select and exclude specific activity data from being used for future recommendations, presented in a timeline format, and marks such data as unavailable for future use, thereby improving recommendation relevance and user control.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If all user-related data is collected and used for recommendations, then the quantity of training data increases, but recommendation accuracy decreases due to uncorrelated data
Solution Approach 1:
The patent extracts and removes uncorrelated data (outliers) from the user activity dataset before training recommendation algorithms. This is achieved through automated detection mechanisms that identify activities unrelated to user preferences, such as purchasing items for others, and excludes them from the training data to improve recommendation accuracy while maintaining sufficient data quantity.
Solution Approach 2:
The patent applies different quality standards to different portions of user data by categorizing activities as correlated or uncorrelated with user preferences. Rather than treating all data uniformly, the system applies local quality filtering to retain only high-quality correlated data for recommendation training, ensuring each data point contributes meaningfully to accuracy.
2Adaptability or versatility
If user feedback is used to retrain algorithms, then user control over recommendations improves, but uncorrelated data continues to affect future recommendations
Solution Approach 1:
The patent performs preliminary action by automatically detecting and removing uncorrelated data before the recommendation algorithm is trained and before user feedback is incorporated. This proactive filtering ensures that uncorrelated data never enters the training pipeline, preventing it from affecting future recommendations even as the system adapts through user feedback.
Solution Approach 2:
The patent incorporates user feedback mechanisms that allow users to indicate the accuracy or relevance of recommendations. This feedback is used to retrain recommendation algorithms, enabling the system to adapt and improve over time while maintaining reliability through the continuous exclusion of uncorrelated data identified by automated detection.
3Ease of operation
If user consent is obtained for data collection, then user privacy control improves, but distinction between consent to collect and consent to use for recommendations is lost
Solution Approach 1:
The patent segments user data consent into distinct categories: consent to collect data and consent to use data for recommendations. This segmentation allows the system to separately manage and control different aspects of data usage, providing users with granular privacy control and preserving the distinction between collection and usage intent through separate consent mechanisms.
Data Source
AI summary
Systems and methods include receipt, from a machine learning system, of a first output based on a first plurality of activity data associated with a user, presentation of the first output to the user, presentation of the first plurality of activity data to the user, receipt of an instruction from the user to designate a first portion of the first plurality of the activity data as unavailable for generation of output, and, in response to the instruction, designation of the first portion of the first plurality of activity data as unavailable for generation of outputs for the user.