Digital Persona Grouping via Activity Hierarchy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional clustering systems and analytics recommendation systems face challenges such as high computational requirements, inflexibility in domain knowledge, and inaccurate digital content recommendations due to reliance on complex machine-learning models and feature engineering processes.
Innovation Solution
The persona group system organizes user activity data into a hierarchy of digital actions, tasks, and workflows, generating vector representations to categorize users into persona groups, enabling faster and more accurate digital recommendations without requiring extensive domain knowledge.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional clustering systems use machine-learning models to perform complex analyses to predict user segments, then user segmentation capability is improved, but processing power consumption and execution time increase inordinately
Solution Approach 1:
The patent extracts and eliminates the complex machine-learning model processing step from the user segmentation pipeline. Instead of using heavy ML models, the system directly analyzes user activity data through simpler computational methods, removing the bottleneck that caused excessive processing power consumption and slow execution while maintaining segmentation capability
Solution Approach 2:
The patent replaces expensive, computationally intensive machine-learning models with cheaper, lighter computational approaches. The system uses straightforward data analysis methods that consume minimal processing resources and execute quickly, achieving the same segmentation goal without the overhead of training and running complex ML models
2Extent of automation
If conventional clustering systems use machine-learning models to form clusters, then cluster formation capability is improved, but processing power consumption increases
Solution Approach 1:
The patent substitutes the mechanical/computational complexity of machine-learning model execution with a simpler data-driven analysis system. The system replaces ML-based automated cluster formation with direct pattern recognition and grouping algorithms that require minimal computational resources, thereby reducing processing power consumption while maintaining automated functionality
Solution Approach 2:
The patent creates a simplified copy or approximation of the ML-based clustering process that achieves similar results with far less computational expense. The system uses lightweight algorithms that replicate the essential clustering functionality without the heavy processing requirements of actual machine-learning models
3Quantity of substance
If conventional analytics recommendation systems conduct extensive surveys to gather user data, then data collection completeness is improved, but time consumption and system complexity increase
Solution Approach 1:
The patent implements preliminary data collection through continuous passive tracking of user activity data in the background, rather than conducting extensive active surveys. The system continuously gathers user interaction information, clicks, and behavior patterns as users naturally engage with the platform, eliminating the need for time-consuming survey administration while maintaining comprehensive data collection
Solution Approach 2:
The system enables users to indirectly provide data through their natural interactions with the platform. As users perform routine tasks and engage with content, the system automatically captures and stores their activity data without requiring users to consciously participate in surveys or data collection processes, thereby reducing time consumption while maintaining data completeness
4Extent of automation
If conventional analytics recommendation systems rely on machine-learning models for content recommendations, then recommendation capability is improved, but accuracy decreases due to faulty inferences
Solution Approach 1:
The patent extracts and removes the faulty machine-learning inference layer from the recommendation system. Instead of relying on ML models that make erroneous predictions, the system directly analyzes user activity data and generates recommendations based on observed patterns and behaviors, eliminating the source of inaccurate inferences while maintaining automated recommendation generation
Solution Approach 2:
The patent introduces an intermediary data analysis layer between raw user data and recommendation output that replaces the faulty ML inference mechanism. This intermediary system processes user activity data through transparent, interpretable analytical methods that accurately reflect user preferences and behaviors, producing more reliable recommendations without the black-box errors of machine-learning models
Data Source
AI summary
The present disclosure relates to systems, non-transitory computer-readable media, and methods that utilize a data-driven approach to organize user-activity data for a user into a hierarchy of digital actions, digital tasks, and digital workflows and categorize a vector representing frequent activities from the hierarchy into a persona group for the user. From this vector representation, the disclosed systems can categorize the vector representation from among a distribution of other vector representations for other users into a persona group for the particular user. Based on at least one of the determined persona group or the vector representation, the disclosed systems can use a nodal graph to determine a digital recommendation that the particular user collaborate with other users or collaborate on a particular project.