Causal Contribution Ranking for Anomaly Attribution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional data analysis systems are inflexible, inaccurate, and inefficient in determining the contributions of dimension values to anomalies, often attributing credit based on face-value representations and requiring extensive computational time.
Innovation Solution
A causal contribution system that determines causal contributions of dimension values to anomalous data by traversing a causal network representing dependencies between dimensions, using simulated interventions and a causal mixture model to generate a ranking of dimension values based on their causal contributions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If conventional data analysis systems use rigid face-value representation to determine dimension value contributions, then the system operation is simple, but the analysis flexibility and accuracy deteriorate
Solution Approach 1:
The patent introduces a causal dependency model as an intermediary between the raw data and the analysis results. This model mediates the relationship by translating face-value representations into causal contributions through learned dependency relationships, enabling flexible analysis while maintaining operational simplicity through automated mediation.
Solution Approach 2:
The system changes the parameters used for analysis from direct face-value representations to causal contribution values derived from the causal dependency model. This parameter transformation enables the system to maintain simple operations while achieving flexible and accurate analysis by working with transformed parameters rather than raw data.
2Productivity
If conventional systems rigidly attribute anomaly contributions based on face-value representations, then the processing is fast, but the measurement precision and accuracy deteriorate
Solution Approach 1:
The causal dependency model is pre-trained using historical data before actual anomaly analysis occurs. This preliminary action stores learned causal relationships in the model, enabling fast and accurate contribution determination during runtime without requiring complex real-time computations, thus resolving the contradiction between speed and precision.
Solution Approach 2:
The system creates a simplified causal dependency model that copies the essential causal relationships from complex real-world data patterns. This model serves as a representative abstraction that enables fast and accurate contribution analysis by working with the simplified model rather than raw complex data.
3Ease of manufacture
If conventional systems use traditional models to determine dimension value contributions, then the implementation is straightforward, but the computational time and efficiency deteriorate
Solution Approach 1:
The causal dependency model is pre-computed and stored during an offline training phase using historical data. This preliminary action consolidates complex computational work into the pre-trained model, enabling fast query responses during runtime without requiring extensive computational time, thus resolving the contradiction between implementation ease and computational efficiency.
Data Source
AI summary
The present disclosure relates to methods, systems, and non-transitory computer-readable media for determining causal contributions of dimension values to anomalous data based on causal effects of such dimension values on the occurrence of other dimension values from interventions performed in a causal graph. For example, the disclosed systems can identify an anomalous dimension value that reflects a threshold change in value between an anomalous time period and a reference time period. The disclosed systems can determine causal effects by traversing a causal network representing dependencies between different dimensions associated with the dimension values. Based on the causal effects, the disclosed systems can determine causal contributions of particular dimension values on the anomalous dimension value. Further, the disclosed systems can generate a causal-contribution ranking of the particular dimension values based on the determined causal contributions.