Automatic Parameter Setting for Time-Series Data Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional data analysis systems require preliminary information and hyperparameters for setting parameters, making it difficult to set suitable parameters for monitoring targets without initial data, especially when preliminary information is lacking.
Innovation Solution
A data analysis device that automatically sets parameters by extracting temporal change and relevance features from time-series data, determining a model based on these features, and using the corresponding parameter setting configuration to set analysis parameters for the monitoring target system.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If preliminary information and hyperparameters are required for parameter setting, then parameter accuracy can be improved, but system complexity and ease of operation deteriorate due to the need for manual input and preliminary data collection
Solution Approach 1:
The system automatically extracts features from time-series data and determines parameters without requiring manual input of preliminary information. The parameter setting unit performs self-service by using the extracted temporal change features and relevance features to automatically determine model type and set appropriate parameters, eliminating the need for operators to manually collect and input preliminary data.
Solution Approach 2:
The system performs preliminary feature extraction and analysis automatically before parameter setting. By extracting temporal change features and relevance features from the time-series data itself, the system prepares the necessary information in advance without requiring external preliminary data collection, thus improving both accuracy and ease of operation.
2Reliability
If preliminary information is required for parameter setting, then parameter reliability can be improved, but loss of time and productivity deteriorate due to additional investigation costs
Solution Approach 1:
The system automatically extracts features from the available time-series data and determines parameters without requiring manual collection of preliminary information. This self-service approach maintains parameter reliability by using the actual monitoring data itself for feature extraction, while eliminating the time loss associated with separate preliminary investigation phases.
Solution Approach 2:
The system extracts necessary features directly from the time-series data by separating temporal change features and relevance features. This extraction process obtains all needed information from the existing data without requiring additional preliminary information collection, thus maintaining reliability while reducing time loss.
3Adaptability or versatility
If manual input of preliminary information is required, then adaptability to different monitoring targets can be improved, but device complexity increases due to multiple input requirements
Solution Approach 1:
The system uses a universal feature extraction approach that works across different monitoring targets. By extracting temporal change features and relevance features from time-series data in a unified manner, the system adapts to different monitoring targets without requiring target-specific preliminary information input, thus maintaining adaptability while reducing complexity.
Solution Approach 2:
The parameter setting unit automatically adapts to different monitoring targets by using the features extracted from their respective time-series data. This self-service adaptation eliminates the need for manual configuration for each target type, maintaining versatility while reducing system complexity.
Data Source
AI summary
A data analysis device executes: acquiring time-series data including data of multiple items from an analysis target system; extracting data of at least, one first item from the time series data; calculating a feature of a temporal change of the data of the first item; extracting data of at least one second item from the time series data; calculating a feature of at least one of a relevance between the data of the first item and the data of the second item and a relevance between the data of the multiple second items; determining a model corresponding to the analysis target system on the basis of the feature of the temporal change and the change of the relevance; and setting the parameter in the analysis target system by using the parameter setting configuration of the model.