Missing Operational Data Restoration via Stochastic Substitution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for processing missing data in operational data, such as event logs, have low restoration rates and are not effectively applicable due to the different properties of operational data compared to numerical observation data.
Innovation Solution
A method and device that identify missing data in operational data by selecting candidate data based on the form defined by events and resources, using a missing table to extract and process candidate data for accurate restoration, leveraging the distribution of the entire data set through stochastic substitution.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional statistical methods (mean substitution, stochastic substitution) are applied to operational data, then the restoration process can be performed, but the restoration rate remains low (about 40-50%)
Solution Approach 1:
The patent segments operational data into four distinct characteristics levels and further divides them into three missing types, creating 12 specific missing cases. This segmentation allows tailored restoration approaches for different data types rather than applying generic statistical methods uniformly, thereby improving restoration rates while managing complexity through structured classification.
Solution Approach 2:
The patent changes the restoration parameters by using event-log-specific methods including candidate data extraction from missing tables, distribution derivation of entire data sets, and stochastic substitution multiple times. These parameter changes adapt the restoration process to operational data characteristics, achieving higher restoration rates compared to conventional statistical methods.
2Adaptability or versatility
If conventional missing data processing technology is applied to operational data, then restoration can be performed, but the restoration rate is low making it difficult to apply to practical use
Solution Approach 1:
The patent applies local quality by selecting candidate data based on the specific form in which resources are defined by events and locations where missing data is identified. Different restoration approaches are applied to different data characteristics levels and missing types, making the method adaptable to operational data while achieving high restoration rates through localized restoration strategies.
Solution Approach 2:
The patent performs preliminary actions by deriving the distribution of the entire data set before substitution, and by pre-processing candidate data extraction from missing tables. These preliminary steps prepare the restoration process to handle operational data effectively, improving both adaptability and restoration rate by establishing proper data distributions and candidate pools before actual restoration.
3Measurement precision
If generic statistical restoration methods are used, then the process is simple, but the restoration accuracy is insufficient for operational data with different properties
Solution Approach 1:
The patent introduces dynamics by performing stochastic substitution multiple times and by adaptively selecting restoration schemes based on data characteristics levels and missing types. The restoration process dynamically adjusts to different operational data properties, achieving high restoration accuracy while managing complexity through structured classification of 12 missing cases.
Solution Approach 2:
The patent implements feedback mechanisms by evaluating restoration results and using distribution derivation from entire data sets to guide subsequent substitutions. The system continuously refines restoration accuracy by comparing restored data against derived distributions and adjusting candidate data selection accordingly, achieving high precision while maintaining manageable complexity through iterative improvement.
Data Source
AI summary
Disclosed are a device and a method for restoring missing operational data. The method for restoring missing operational data includes determining whether missing data is present in a first event defining operational data or a first resource constituting the operational data, extracting candidate data from a missing table, depending on a form in which the first resource is defined by the first event and a location where the missing data is identified, and processing the candidate data to restore the missing data, based on a predetermined restoration scheme.