Prerequisite Relationship Extraction Using Surprise and Causality
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods struggle to extract unexpected preceding relationships between time-series variables that are difficult to conceive by human sense and are not useful for future prediction.
Innovation Solution
A preceding relationship extraction device and method that calculates a degree of preceding, semantic similarity, and surprise degree, and performs causality testing to identify and present significant relationships between time-series data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional methods (CCF or VAR) are used to extract preceding relationships, then the extraction process is simple, but only easily conceivable relationships are extracted and unexpected relationships are missed
Solution Approach 1:
The extraction process is divided into distinct functional modules: a preceding degree calculation unit that quantifies temporal precedence, a similarity calculation unit that measures semantic relatedness, and a surprise degree calculation unit that identifies unexpected relationships. This segmentation allows each module to specialize in one aspect of relationship detection, improving overall precision while maintaining manageable system complexity through modular design.
Solution Approach 2:
The system introduces new parameters (preceding degree, similarity degree, surprise degree) that transform the traditional binary presence/absence detection into a multi-dimensional continuous assessment. By calculating these parameters and using them to filter and rank relationships, the system can precisely identify unexpected preceding relationships that conventional methods miss, while the parameterized approach keeps the system structure organized and manageable.
2Reliability
If conventional methods are used, then the system is easy to implement, but relationships useful for future prediction cannot be extracted
Solution Approach 1:
The surprise degree acts as an intermediary metric that bridges the gap between simple temporal correlation and complex causality prediction. By first identifying relationships with high preceding degree and high surprise degree (indicating unexpected but strong temporal patterns), and then applying causality testing only to these candidates, the system reliably identifies prediction-useful relationships while avoiding the complexity of comprehensive causality testing across all variable pairs.
3Measurement precision
If all possible relationships are analyzed, then comprehensive coverage is achieved, but computational resources are wasted on obvious relationships
Solution Approach 1:
Instead of performing exhaustive analysis on all variable pairs, the system applies partial action by first calculating preceding degree and similarity for all pairs, then using the surprise degree metric to selectively identify only those relationships worth further investigation. This approach achieves high accuracy in identifying unexpected relationships while significantly improving computational efficiency by avoiding wasted resources on obvious or irrelevant relationships.
Data Source
AI summary
The present invention includes: a preceding degree calculation unit (1) that calculates a degree of preceding of time series data xj of an item j with respect to time series data xi of an item i from a plurality of pieces of data; a similarity calculation unit (2) that calculates a semantic similarity between the time series data xi and the time series data xj; a surprise degree calculation unit (3) that calculates a degree of surprise indicating surprise of combining the item i and the item j on the basis of the degree of preceding and the semantic similarity; a causality testing unit (4) that tests causality of the item i and the item j; and a presentation unit (5) that presents the degree of surprise and presence or absence of the causality.