Identifying Recurrent Causal Sequences in Temporal Datasets
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional methods for identifying recurrent sequences in temporal datasets primarily focus on temporal relationships, which may lead to overlooking significant causal patterns, insufficient for understanding the dynamics of complex systems.
Innovation Solution
A computer-implemented method and system that computes the degree of causal relatedness between data points in multiple datasets, constructs directed acyclic graphs based on this relation, and identifies recurrent causal sequences by constructing a second-order graph from these directed acyclic graphs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional methods focus on temporal relationships to identify recurrent sequences, then temporal pattern recognition is improved, but causal pattern identification deteriorates
Solution Approach 1:
The patent segments the analysis into two independent parts: temporal relationship analysis and causal relationship analysis. Each part is handled separately using appropriate methods (temporal convolution for temporal patterns, causal discovery algorithms for causal patterns), then the results are integrated. This segmentation allows each analysis to optimize for its specific purpose without compromising the other.
Solution Approach 2:
The patent transitions from analyzing data in a single temporal dimension to analyzing it in both temporal and causal dimensions simultaneously. By constructing a multi-dimensional representation that includes time sequences and causal graphs, the system can identify patterns that exist in either dimension or their interaction, thereby recovering causal information that was previously lost in single-dimension temporal analysis.
2Productivity
If temporal relationships are used to identify recurrent sequences, then sequence detection is improved, but understanding of complex system dynamics deteriorates
Solution Approach 1:
The patent merges temporal sequence detection with causal relationship analysis into a unified framework. The system combines temporal convolution networks for sequence detection with causal discovery algorithms to identify causal structures. By merging these complementary approaches, the system achieves both efficient sequence detection and reliable understanding of system dynamics, as the causal analysis provides contextual interpretation of detected sequences.
Solution Approach 2:
The patent introduces causal graphs as an intermediary structure that mediates between raw temporal data and high-level system dynamics understanding. The causal graph serves as an intermediate representation that captures causal relationships among variables, allowing the system to interpret temporal sequences in terms of causal mechanisms, thereby bridging the gap between efficient detection and reliable understanding.
Data Source
AI summary
A computer-implemented method for identifying recurrent causal sequences in datasets is provided. The method comprises receiving at least first and second datasets each comprising a plurality of data points and computing a degree of causal relatedness for all pairs of the data points of the first and second datasets. The method comprises constructing first and second directed acyclic graphs on top of the respective first and second datasets. The method comprises constructing a second order graph using the first and the second directed acyclic graphs and identifying recurrent sequences of patterns in the first and second directed acyclic graphs based on the second order graph.