Time Series Data Aggregation and Clustering for Event Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for analyzing time series data struggle to effectively analyze multiple events by considering complex relationships among multiple pieces of time series data, making it difficult to analyze the appearance of each event in a comprehensive manner.
Innovation Solution
A data processing apparatus that aggregates multiple types of sampling time series data, classifies the aggregated data into clusters, and generates time series appearance data for each cluster, allowing for the analysis of multiple events by plotting the data in a multi-dimensional space and calculating appearance probabilities based on cluster membership.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If multiple types of sampling time series data are aggregated and classified into clusters to generate appearance data for each cluster, then the ability to analyze multiple events by considering complex relationships among multiple time series data is improved, but the device complexity and computational requirements increase
Solution Approach 1:
The patent segments the complex analysis task into distinct processing steps: data aggregation step (combining N types of sampling time series data into M types of classification time series data where N>M), classification step (grouping M types into clusters), and appearance data generation step (creating time series appearance data for each cluster). This segmentation allows the system to handle multiple events systematically while managing complexity through structured processing stages.
Solution Approach 2:
The patent transforms N-dimensional sampling time series data into M-dimensional classification time series data (where N>M) through aggregation, effectively reducing dimensionality. This dimensional transformation simplifies the data structure while preserving essential relationships, enabling the system to analyze complex multi-event patterns without being overwhelmed by high-dimensional complexity.
2Device complexity
If N types of sampling time series data are aggregated into M types of classification time series data through principal component analysis or factor analysis, then the essential patterns in the data are preserved while reducing complexity, but information loss may occur during the aggregation process
Solution Approach 1:
The patent employs principal component analysis or factor analysis to transform the data, changing the parameters from N original sampling time series to M classification time series. These mathematical transformations identify and preserve the dominant patterns and relationships in the data, retaining essential information while reducing dimensionality. The analysis methods ensure that the most significant variations in the original data are captured in the aggregated classification data.
Data Source
AI summary
A data processing apparatus includes an arithmetic processing unit that executes a data aggregation step of aggregating N types (N≥3) of sampling time series data to acquire M types (M≥2 and N>M) of classification time series data, a classification step of classifying the M types of classification time series data into a plurality of clusters, and an appearance data generation step of generating time series appearance data for each cluster.