Chart Indexing via Sliding Window for Pattern Search
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Comparing and searching for similar patterns in line charts across different time intervals and scales is challenging due to varying resolutions and units, making it difficult to visualize trends and predict events like device overheating.
Innovation Solution
The implementation of a sliding window algorithm for indexing charts, which creates overlapping sub-charts that can be stored and searched efficiently, allowing for the comparison and normalization of chart features across different resolutions and scales, enabling the detection of similar patterns.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If line charts are stored with different time intervals and scales, then more detailed data can be preserved, but comparing and searching for similar patterns becomes more difficult
Solution Approach 1:
The patent applies segmentation by dividing charts into overlapping sub-charts of fixed size using a sliding window algorithm. This breaks down complex charts with varying resolutions into uniform, comparable segments that can be indexed and searched efficiently while preserving detailed local patterns.
Solution Approach 2:
The patent transforms chart data by normalizing sub-chart features to a standard scale. This parameter transformation allows sub-charts from different original charts with different time intervals and Y-axis scales to be compared directly, resolving the comparison difficulty while maintaining the original data's detailed information.
2Difficulty of detecting and measuring
If charts are normalized to enable comparison, then pattern detection improves, but the original data resolution and scale information may be lost
Solution Approach 1:
The patent performs preliminary normalization on sub-chart features during the indexing phase, before any search or comparison operations. This preliminary transformation enables efficient pattern detection while the system retains the ability to retrieve and display original chart data with its full resolution and scale information when needed.
Solution Approach 2:
The patent creates normalized copies of chart sub-sections for indexing and search purposes, while the original charts with their full resolution and scale information are preserved separately. This copying approach allows pattern detection on normalized data without losing the original detailed information.
3Productivity
If a sliding window algorithm is used to create sub-charts, then search efficiency improves, but the complexity of the indexing process increases
Solution Approach 1:
The sliding window algorithm systematically segments charts into uniform sub-charts with fixed dimensions and overlapping boundaries. This structured segmentation, while adding indexing complexity, enables efficient search operations by creating a standardized index that can be quickly queried for pattern matching.
Solution Approach 2:
The patent creates a universal indexing structure that handles multiple chart types, resolutions, and scales through a single normalized sub-chart representation. This universal approach increases the sophistication of the indexing process but enables the system to efficiently search across diverse chart data using the same indexed structure.
Data Source
AI summary
Approaches for indexing and comparing charts are described. A system can receive one or more charts, which may include depictions of signals, and index portions of a chart using a sliding window algorithm. Subsequently, a system can receive a query that can be compared to the indexed portions of one or more charts. After a comparison, the most similar portions of the compared charts are provided based on a nearest neighbour search.