Dynamic Content Frequency Control via Machine Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing frequency cap rules for controlling content item transmission frequency are inflexible, fail to adapt to changes in user behavior, and do not account for individual user responses, leading to reduced effectiveness and user inattention.
Innovation Solution
A machine-learning based system that adjusts predicted user interaction rates by learning from multiple features, including impression counts, to dynamically control content item frequency without relying on traditional frequency cap rules, thereby improving user experience and increasing content item selection depth.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If traditional frequency cap rules are used to control content item transmission frequency, then content item frequency can be controlled, but the system fails to adapt to changes in user behavior and supply and demand
Solution Approach 1:
The patent applies dynamics by transitioning from static frequency cap rules to a dynamic machine learning model that continuously adapts to changing user behavior and supply-demand conditions. The model learns from historical data and adjusts frequency predictions in real-time, making the system responsive to temporal changes while maintaining reliable frequency control through learned patterns.
Solution Approach 2:
The patent changes the parameter representation from fixed cap thresholds to continuous probability distributions learned by the machine learning model. By modeling frequency as a probabilistic parameter rather than a hard constraint, the system can adapt to varying user preferences and content characteristics while maintaining effective frequency control through learned parameter adjustments.
2Ease of operation
If frequency cap rules are applied globally, then frequency control is simplified, but individual user responses towards repeated content items are ignored
Solution Approach 1:
The patent applies local quality by transitioning from global frequency cap rules to user-specific frequency predictions generated by the machine learning model. Each user receives personalized frequency control based on their individual response patterns, while the overall system maintains operational simplicity through automated model-based decisions rather than manual configuration for each user.
Solution Approach 2:
The machine learning model performs self-service by automatically learning user-specific frequency preferences from historical data without requiring manual configuration. The system serves itself by generating personalized frequency predictions through learned patterns, eliminating the need for complex manual rule setup while capturing individual user responses.
3Device complexity
If frequency cap rules are reset after a fixed period, then implementation is simplified, but user feedback towards repeated content items outside that period cannot be captured
Solution Approach 1:
The patent applies continuity by replacing periodic reset mechanisms with continuous learning through the machine learning model. The model continuously processes user feedback and historical data without interruption, maintaining an ongoing understanding of user preferences that captures feedback across all time periods rather than losing information at reset boundaries.
Solution Approach 2:
The machine learning model performs preliminary action by learning from historical user feedback before making frequency predictions. By pre-learning patterns from past interactions, the system captures and utilizes user feedback information proactively rather than losing it at reset points, while maintaining implementation simplicity through automated model training.
4Reliability
If an f-cap rule is applied, then content item frequency is controlled, but the depth of content item selection event is unnecessarily reduced
Solution Approach 1:
The patent applies dynamics by replacing static f-cap rules with dynamic machine learning-based frequency predictions that are integrated into the content selection event. This allows frequency control to be applied adaptively based on learned user preferences, maintaining control effectiveness while preserving selection event depth through probabilistic rather than hard constraint-based decisions.
Data Source
AI summary
Techniques for controlling item frequency using machine learning are provides. In one technique, two prediction models are trained: one based on interaction history of multiple content items by multiple entities and the other based on predicted interaction rates and an impression count for each of multiple content items. In response to a request, a particular entity associated with the request is identified and multiple candidate content items are identified. For each identified candidate content item, the first prediction model is used to determine a predicted interaction rate, an impression count of the candidate content item is determined with respect to the particular entity, the second prediction model is used to generate an adjustment based on the impression count, and an adjusted entity interaction rate is generated based on the predicted interaction rate and the adjustment. A particular candidate content item is selected based on the generated adjusted entity interaction rates.