Joint Predictive Model for Ad Targeting Segments
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current ad targeting methods face challenges in predictive power and scalability due to separate modeling of user segments, inability to consider interactions among segments, and inefficiencies in hardware and computing resources, especially in a cookieless environment.
Innovation Solution
Implementing a joint predictive model using extreme multi-label classification and Factorization Machine models to simultaneously predict conversion probabilities for multiple audience segments, leveraging contextual data without relying on user identity tracking, and using a single model to estimate performance across all segments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If separate prediction models are used for each targeting segment, then the prediction can be performed for individual segments, but the predictive power is reduced and hardware and computing times are wasted
Solution Approach 1:
The patent merges multiple separate prediction models into a single unified prediction model that processes multiple targeting segments simultaneously. This consolidation maintains predictive accuracy while reducing computational overhead and hardware requirements by sharing common processing components across all segments.
Solution Approach 2:
The unified prediction model is designed to perform multiple functions by predicting outcomes for all targeting segments within a single model structure. This multi-functional approach eliminates the need for separate specialized models for each segment, optimizing resource utilization while maintaining segment-specific prediction capabilities.
2Adaptability or versatility
If separate prediction models are used for each targeting segment, then individual segment prediction is possible, but scalability is reduced
Solution Approach 1:
By combining multiple segment predictions into a unified model, the system achieves better scalability as the model can accommodate additional segments without requiring proportional increases in computational infrastructure. The shared architecture allows efficient handling of growing segment numbers.
Solution Approach 2:
The unified model uses a single shared prediction structure that can be applied across all segments, eliminating the need to create and maintain separate model copies for each segment. This approach significantly improves scalability while maintaining segment-specific prediction accuracy.
3Reliability
If separate prediction models are used for each targeting segment, then segment-specific prediction can be performed, but hardware and computing times are wasted
Solution Approach 1:
The patent merges multiple segment predictions into a single unified model that shares computational resources across all segments. This consolidation maintains prediction accuracy for each segment while significantly reducing the total computing power, memory, and energy required compared to maintaining separate models for each segment.
Data Source
AI summary
This teaching relates to predictive targeting. Training data are obtained with pairs of data. Each pair includes an ad opportunity context corresponding to an ad served to a plurality of audiences and a label vector having a plurality of labels, each of which indicates a reaction, with respect to the ad served, of a corresponding one of the audiences in the ad opportunity context. Based on the training data, model parameters of a joint predictive model are learned via machine learning based on an initialized model with initial model parameters by minimizing a loss in an iterative process. The learned joint predictive model is to be used to map an input context of an ad opportunity to an output label vector having a plurality of probabilities, each of which predicts a likelihood of a reaction of a corresponding one of the audiences to the input context of the ad opportunity.