Ensemble Deep Learning for Pre-Deployment Campaign Anomaly Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing digital media campaign configurations often contain erroneous or unusual settings that lead to unsuccessful campaigns, resulting in resource wastage and increased troubleshooting efforts.
Innovation Solution
An ensemble of deep learning models, including deep belief networks, generative adversarial networks, and one-class support-vector data description models, is used to identify anomalous campaign settings before deployment, providing error messages to correct these settings.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional anomaly detection methods are used for media campaign configuration, then the system is simpler to implement, but the detection accuracy and reliability of identifying erroneous settings is insufficient
Solution Approach 1:
The anomaly detection system is segmented into multiple specialized machine learning models, each trained to detect specific types of anomalies in media campaign configurations. This segmentation allows each model to focus on particular patterns, improving overall detection accuracy while maintaining manageable complexity through modular architecture.
Solution Approach 2:
Multiple machine learning models are merged into an ensemble system that combines their predictions through weighted voting. This merging leverages the strengths of different models to achieve superior anomaly detection accuracy compared to individual models, resolving the contradiction between complexity and precision by distributing the detection burden across multiple specialized components.
2Reliability
If multiple machine learning models are used to improve anomaly detection, then the detection reliability is enhanced, but the computational resources and time required increase
Solution Approach 1:
The machine learning models are pre-trained on historical media campaign data before deployment. This preliminary action allows the models to learn anomaly patterns in advance, enabling rapid real-time detection during actual campaign configuration without requiring extensive computation during the detection phase itself.
Solution Approach 2:
The system implements feedback mechanisms where model predictions and detection results are continuously evaluated and used to refine the weighted voting process. This feedback loop improves detection reliability over time while optimizing the balance between multiple models to reduce unnecessary computational overhead.
3Productivity
If comprehensive anomaly detection is performed on all campaign settings, then the success rate of media campaigns is improved, but the computational resources and financial cost increase
Solution Approach 1:
The system extracts and flags only the specific anomalous configuration settings from the complete media campaign configuration, rather than processing all settings uniformly. This extraction approach identifies only the problematic elements that require attention, reducing unnecessary computational resources while maintaining comprehensive detection coverage for critical parameters.
Solution Approach 2:
The system dynamically adjusts the detection sensitivity and model selection based on the specific campaign type and configuration parameters. By changing detection parameters adaptively, the system optimizes the balance between thorough anomaly detection and computational resource consumption, preventing waste while ensuring campaign success.
4Ease of operation
If detailed error messages are provided to guide users in correcting settings, then the ease of operation is improved, but the complexity of the error reporting system increases
Solution Approach 1:
The system acts as an intermediary between the complex machine learning models and the end user by translating model predictions into user-friendly error messages. This intermediary layer simplifies the interaction for users while managing the underlying model complexity, providing actionable guidance without exposing users to the complexity of the ensemble modeling system.
Data Source
AI summary
The subject technology detects anomalies in media campaign configuration settings. The anomaly detection system may leverage one or more deep learning models to detect anomalies and identify particular configuration settings that contribute to the detected anomalies. In various embodiments, two or more of the deep learning models may be combined into an ensemble model that boosts the accuracy of anomaly predictions made by the anomaly detection system. The anomaly detection system may review the configuration settings of media campaigns during the configuration process and before the media campaigns run on a publication system in order to reduce the amount of unsuccessful campaigns and minimize the amount of wasted resources spent on running campaigns that have a low likelihood of achieving user defined goals.