Numeric Data Stream Change Point Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing techniques fail to effectively compare numeric data streams with discrete event occurrences due to the incompatibility between numeric and discrete data types.
Innovation Solution
The method involves converting numeric data streams into equivalent discrete event occurrences by detecting change points using Shewhart control charts and cumulative sum values, allowing for comparison with discrete event occurrences.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If existing techniques are used to compare numeric data streams with discrete event occurrences, then data correlation analysis can be performed on comparable data types, but numeric data and discrete data cannot be compared due to type incompatibility
Solution Approach 1:
The patent transforms numeric data streams by detecting change points and converting them into discrete event occurrences with specific parameters (event type, timestamp, magnitude). This parameter transformation enables numeric data to be expressed in the discrete domain, making it compatible with discrete event occurrences for correlation analysis while preserving the essential characteristics of the original numeric data through structured parameter representation
2Adaptability or versatility
If numeric data streams are converted to discrete events for comparison, then correlation analysis between numeric and discrete data becomes possible, but the conversion process requires detecting change points which adds computational complexity
Solution Approach 1:
The patent introduces change point detection as an intermediary process that bridges numeric data streams and discrete event occurrences. This intermediary mechanism identifies significant transitions in numeric data and transforms them into standardized discrete events, providing a systematic conversion pathway that manages computational complexity through established statistical methods rather than direct complex mapping
Solution Approach 2:
The patent segments numeric data streams by identifying and isolating change points that represent significant transitions. By dividing the continuous numeric data into discrete segments marked by change points, the system converts a complex continuous analysis problem into simpler discrete event analysis, reducing overall computational complexity while maintaining analytical accuracy
3Extent of automation
If change point detection is used to convert numeric data streams into discrete events, then automatic discovery of correlations becomes possible, but the search space for numeric data correlation is much larger than for discrete data
Solution Approach 1:
The patent performs preliminary change point detection on numeric data streams before conducting correlation analysis. By pre-processing the numeric data to identify and mark significant change points, the system reduces the search space for subsequent correlation analysis, transforming a large continuous search problem into a smaller discrete search problem that can be solved more efficiently with less computation time
Data Source
AI summary
Embodiments of the present invention relate to a method of data analysis, comprising comparing a data point in a data stream and a previous data point in the data stream to identify a change direction of the data point based on a difference between the data point and the previous data point, comparing the change direction with a previous change direction to identify whether a trend change occurred for the data point, and identifying a change point that comprises a discrete data value in the data stream based on the trend change. The method also comprises comparing the change point with other data.