Propensity Model for Online Learning Target Identification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing digital content delivery systems face challenges in predicting user affinity for broad categories of digital content, such as online learning programs, where demographic correlations are weak and rule-based systems become complex and inefficient, and machine-learning models struggle when features of digital content are unknown.
Innovation Solution
A machine-learned propensity model that identifies users with traits like 'perpetual student' or 'life-long learner' by correlating digital feature data without content-specific data, using user-specific digital signals and similarity features from aggregate user data to predict engagement with broad categories of digital content.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If rule-based systems are used to predict user affinity for digital content, then demographic correlations can be established, but the systems become complex and inefficient when dealing with broad content categories
Solution Approach 1:
The patent replaces the mechanical rule-based system with a machine learning model that automatically learns patterns from user interaction data. Instead of manually defining complex rules to predict user affinity for broad content categories, the system uses algorithms to identify correlations between user behaviors and content engagement, thereby reducing rule set complexity while maintaining or improving prediction accuracy.
Solution Approach 2:
The patent changes the parameters used for prediction from static demographic data to dynamic user interaction features. By focusing on behavioral signals such as click patterns, time spent on content, and interaction history, the system adapts to broad content categories without requiring complex demographic rules, thus simplifying the overall system while improving affinity measurement.
2Measurement precision
If machine-learning models are used to predict user engagement, then complex patterns can be identified, but the models struggle when features of digital content are unknown
Solution Approach 1:
The patent creates a universal machine learning model that can handle both specific and broad content categories. The model is designed to work with any digital content by learning from general user interaction patterns rather than requiring content-specific features. This multi-functionality allows the same model to predict engagement across diverse content types, from news articles to videos, even when content features are unknown or vary significantly.
Solution Approach 2:
The system performs preliminary learning by training on large datasets of user interactions before deployment. This preliminary action allows the model to pre-learn general patterns of user behavior and content engagement, enabling it to make accurate predictions even for new or unknown content features. The model is prepared in advance to handle various content types without requiring specific feature knowledge at prediction time.
3Productivity
If digital content delivery is targeted to specific user groups, then click-through rates can be improved, but handling non-mutually exclusive user groups becomes challenging
Solution Approach 1:
The patent implements a self-service targeting system where the machine learning model automatically determines user group assignments based on individual user profiles and content characteristics. Instead of requiring manual configuration of user groups and their relationships, the system autonomously calculates propensity scores and makes targeting decisions, thereby improving click-through rates while eliminating the complexity of managing non-mutually exclusive groups.
Solution Approach 2:
The patent introduces dynamic user group assignments where users can belong to multiple overlapping segments simultaneously based on their current behavior and context. The system dynamically adjusts user group memberships and targeting strategies in real-time, allowing flexible handling of non-mutually exclusive groups. This dynamic approach enables the system to optimize click-through rates by adapting to changing user preferences without the rigidity of fixed, mutually exclusive segments.
Data Source
AI summary
In an embodiment, the disclosed technologies include determining a digital identifier, computing, using aggregate digital event data obtained from at least one computing device, digital feature data relating to the digital identifier, inputting the digital feature data relating to the digital identifier into a digital model that has machine-learned correlations between digital feature data and digital propensity prediction values, receiving, from the digital model, a predicted propensity value associated with the digital identifier, determining a propensity score based on the predicted propensity value, causing a digital content item to be displayed on a user interface of a network-based software application associated with the digital identifier if the propensity score satisfies a propensity criterion.