Ensemble Anomaly Detection for Media Campaign Configuration
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing digital media campaign systems lack effective anomaly detection during configuration, leading to erroneous or unsuccessful campaigns due to inadequate protection against unusual or unlikely settings, resulting in resource wastage and increased troubleshooting.
Innovation Solution
An ensemble of deep learning models, including deep belief networks, generative adversarial networks, and one-class support-vector data description models, analyzes campaign settings jointly to identify complex anomalies and flag problematic configurations, providing error messages for correction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional anomaly detection methods are used for campaign configuration, then the system is simpler and faster to implement, but the detection accuracy is insufficient leading to erroneous campaigns
Solution Approach 1:
The anomaly detection task is segmented into multiple specialized models: a deep belief network for general anomaly detection, a generative adversarial network for detecting complex non-linear patterns, and a one-class support vector machine for identifying deviations from normal patterns. Each model segment handles specific aspects of anomaly detection, improving overall accuracy while maintaining manageable complexity through modular architecture.
Solution Approach 2:
The system employs a composite ensemble model that integrates multiple different machine learning algorithms (deep belief networks, generative adversarial networks, and one-class support vector machines). This composite approach combines the strengths of different model types to achieve superior detection accuracy that neither individual model could achieve alone, analogous to using composite materials to achieve superior physical properties.
2Reliability
If comprehensive anomaly detection is implemented, then campaign success rate improves, but computational resources and processing time increase
Solution Approach 1:
The ensemble model performs anomaly detection during the campaign configuration phase, before campaigns are deployed. By conducting comprehensive analysis preliminarily, the system identifies and flags problematic configurations early, preventing unsuccessful campaigns from consuming computational resources during execution. This preliminary action ensures high reliability while optimizing resource usage during actual campaign operations.
3Measurement precision
If multiple machine learning models are combined for ensemble detection, then detection precision increases, but system complexity and training difficulty increase
Solution Approach 1:
The ensemble system is segmented into three distinct but complementary models, each with a specific strength: deep belief networks for hierarchical feature extraction, generative adversarial networks for complex pattern generation, and one-class support vector machines for normal pattern definition. This segmentation allows each model to be trained and optimized independently, reducing the complexity of the overall training process while maintaining high detection precision through their combined output.
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.