Event Prediction via Temporal Pattern Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current event classification algorithms face challenges in predicting infrequent but high-impact events due to limited data samples, uneven inter-arrival times, and inefficient storage and search methods, particularly in identifying temporal patterns that predict rare event sub-sequences effectively.
Innovation Solution
A method that characterizes and predicts target events by defining and quantizing measurable events into classes, analyzing temporal series using a sliding time window, and optimizing prediction rules with a fitness function, employing genetic algorithms to search for optimal system parameters for efficient classification and prediction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If event classification algorithms use discriminant description strategy to separate class regions from data, then classification boundaries can be established, but the method is ineffective for rare events with limited data samples
Solution Approach 1:
The patent applies preliminary action by pre-characterizing target events through template matching before classification. Event templates are constructed in advance based on historical data patterns, allowing the system to identify rare events by matching observed events against these pre-defined templates rather than relying solely on discriminant boundaries that require大量 training data.
Solution Approach 2:
The patent uses copying by creating event templates that represent typical patterns of target events. These templates serve as copies or prototypes of rare events, allowing the system to detect and classify rare events by comparing new events against these template copies, thereby overcoming the limitation of having few actual data samples of the target class.
2Adaptability or versatility
If algorithms mine for more than just target events to establish classification rules, then more comprehensive rules can be generated, but storage and search efficiency deteriorates
Solution Approach 1:
The patent applies taking out by extracting only the essential characteristics of target events into compact templates. Instead of storing and processing all event data comprehensively, the system extracts key features that define target events and stores them in condensed template form, enabling efficient storage and rapid search while maintaining the ability to identify rare events.
Solution Approach 2:
The patent uses segmentation by dividing event data into distinct templates based on event types and characteristics. Each template represents a segmented view of specific event patterns, allowing the system to efficiently store and search for target events by comparing against relevant template segments rather than processing entire event datasets.
3Device complexity
If fixed time windows are used for event analysis, then analysis complexity is reduced, but the ability to capture uneven inter-arrival times of rare events is lost
Solution Approach 1:
The patent applies dynamics by using adaptive time windows that adjust their size and positioning based on the characteristics of the events being analyzed. Rather than using fixed time windows, the system dynamically adapts the analysis window to capture the uneven inter-arrival times of rare events, improving detection accuracy while managing complexity through automated adaptation.
4Adaptability or versatility
If manual trial and error is used to estimate time window size, then flexibility in parameter selection is achieved, but time consumption and automation level increase
Solution Approach 1:
The patent applies self-service by implementing automated algorithms that select and optimize time window parameters without requiring manual trial and error. The system uses historical event data to automatically determine appropriate time window sizes and configurations, enabling the method to adapt to different event types while eliminating time-consuming manual parameter tuning and increasing automation.
Data Source
AI summary
A method for characterizing, detecting and predicting an event of interest, a target event, based on temporal patterns useful for predicting a probable occurrence of the target event is disclosed. Measurable events and their features are defined and quantized into event classes. Temporal series of the event classes are analyzed, and preliminary prediction rules established by analyzing temporal patterns of the event classes that precede an occurrence of the target event using a sliding time window. The quality of the preliminary prediction rules is evaluated and parameters thereof are optimized by using a defined fitness function, thereby defining finalized prediction rules. The finalized prediction rules are then made available for application on temporal series of the event classes to forecast a probable occurrence of the target event.