Seasonal Pattern Detection in Time Series Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing time series forecasting models, such as Holt-Winters, primarily focus on generating forecasts and provide limited information on seasonal patterns, making it difficult for end users to interpret and utilize seasonal data outside of the forecasting context, and fail to effectively handle both dense and sparse patterns.
Innovation Solution
The development of techniques to detect, classify, and summarize seasonal patterns within time series data by decomposing it into dense and sparse signals, allowing for independent analysis and classification, and generating summaries that identify recurrent highs and lows across different seasonal classes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If Holt-Winters forecasting algorithm is used to generate forecasts, then forecasting accuracy is improved, but interpretability of seasonal patterns deteriorates
Solution Approach 1:
The patent segments the time series data into dense signal and sparse signal components, allowing separate analysis and interpretation of different seasonal patterns. This segmentation enables the system to maintain forecasting accuracy while providing interpretable seasonal patterns by analyzing dense and sparse components independently.
Solution Approach 2:
The patent introduces seasonal pattern instances as an intermediary representation that bridges the gap between the Holt-Winters forecasting model and interpretable seasonal patterns. These instances serve as a mediator that captures seasonal information in a form that can be independently analyzed and interpreted outside the forecasting context.
2Productivity
If traditional Holt-Winters method is applied, then forecasting capability is improved, but anomaly detection capability deteriorates
Solution Approach 1:
By segmenting the time series into dense and sparse signals, the system can use the sparse signal component specifically for anomaly detection while maintaining the dense signal for forecasting. This segmentation allows the system to leverage seasonal pattern instances for reliable anomaly detection independent of forecasting performance.
Solution Approach 2:
The patent extracts seasonal pattern instances from the time series data that can be used for anomaly detection purposes. By taking out and independently analyzing these seasonal instances, the system gains anomaly detection capability without compromising the forecasting functionality of the Holt-Winters model.
3Productivity
If seasonal indices are implemented as internal structures within forecasting models, then forecasting performance is improved, but ease of operation deteriorates
Solution Approach 1:
The patent introduces seasonal pattern instances as an intermediary that translates the internal seasonal indices of the Holt-Winters model into an externally interpretable format. These instances serve as a bridge that maintains the mathematical integrity of the forecasting model while providing user-friendly, interpretable seasonal patterns that can be independently analyzed.
Solution Approach 2:
The system creates a copy of the seasonal information in the form of seasonal pattern instances that can be independently analyzed and interpreted. This copying allows the seasonal data to be examined outside the forecasting model context, improving ease of operation and interpretability without affecting the forecasting performance.
4Productivity
If time series data is analyzed as a whole, then comprehensive forecasting is improved, but handling of dense and sparse patterns deteriorates
Solution Approach 1:
The patent segments the time series data into dense signal and sparse signal components, enabling precise handling of different pattern types. This segmentation allows the system to apply appropriate analysis methods to each component, improving the measurement precision for detecting and characterizing seasonal patterns while maintaining comprehensive forecasting capability.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Techniques are described for characterizing and summarizing seasonal patterns detected within a time series. According to an embodiment, a set of time series data is analyzed to identify a plurality of instances of a season, where each instance corresponds to a respective sub-period within the season. A first set of instances from the plurality of instances are associated with a particular class of seasonal pattern. After classifying the first set of instances, a second set of instances may remain unclassified or otherwise may not be associated with the particular class of seasonal pattern. Based on the first and second set of instances, a summary may be generated that identifies one or more stretches of time that are associated with the particular class of seasonal pattern. The one or more stretches of time may span at least one sub-period corresponding to at least one instance in the second set of instances.