Machine Learning Models Predict Offsite User Interactions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Collecting and processing user data for offsite users in content delivery systems is resource-intensive and leads to delays in serving relevant content, as onsite and offsite user interactions differ significantly.
Innovation Solution
Implementing a machine-learned model that predicts offsite user interaction rates using a combination of onsite and offsite features, where a first model generates entity-campaign interaction rates and a second model is trained with these rates to select campaigns based on predicted interaction scores, minimizing latency and improving content relevance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If user data is collected and processed for offsite users in real-time, then content relevance is improved, but system latency and resource consumption increase
Solution Approach 1:
The system performs preliminary actions by training offline machine learning models using historical onsite user data before runtime. These pre-trained models capture user preferences and interaction patterns in advance, enabling fast predictions during runtime without real-time data collection and processing delays
Solution Approach 2:
The system creates a copy of the onsite user interaction patterns by training offline models that replicate onsite behavior. These model copies can predict offsite user interactions without requiring actual real-time offsite data collection, thus reducing latency while maintaining prediction accuracy
2Measurement precision
If comprehensive user data is collected from offsite third-party systems, then prediction accuracy is improved, but data scatterness and processing complexity increase
Solution Approach 1:
The system introduces an intermediary mechanism by using trained machine learning models that act as mediators between raw user data and predictions. These models aggregate and process dispersed user interaction patterns into unified prediction outputs, reducing data scatterness and simplifying the complexity of handling multi-source offsite data
3Measurement precision
If real-time user data processing is performed for offsite content delivery, then content selection accuracy is improved, but resource consumption and latency increase
Solution Approach 1:
The system performs content selection preparation in advance by training machine learning models offline using historical data. During runtime, the pre-trained models enable fast content selection without real-time processing overhead, thus improving content delivery efficiency while maintaining selection accuracy
Solution Approach 2:
The system replaces the mechanical approach of real-time data collection and processing with a computational approach using pre-trained machine learning models. This substitution enables fast predictions during runtime without the resource-intensive real-time processing operations
Data Source
AI summary
Techniques for predicting an offsite entity interaction rate are provided. One approach involves using a first machine-learned model that includes a first plurality of features that correspond to entity and campaign attributes. The approach also involves training a second machine-learned model that includes a second plurality of features that includes a particular feature corresponding to predicted entity interaction rates. Thus, output of the first machine-learned model is input to the second machine-learned model. The second machine-learned model includes multiple weights that include a particular weight for the particular feature. A content request is received and a set of campaigns is identified based on an entity identifier associated with the content request. Scores are generated based on the first and second machine-learned models. Based on the scores, a campaign is selected and campaign data associated with the campaign is transmitted over a computer network.