Learning Data Model Configuration for Hierarchical Data Granularity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies struggle to efficiently generate high-quality and quantity learning data for machine learning models by adjusting the degree of abstraction and detail in data items, particularly when integrating databases from different fields, leading to increased processing loads and inefficient data utilization.
Innovation Solution
A method and apparatus that utilize a filter to designate the degree of abstraction or detail for each data item, sorting them into objective and explanatory variables, and generate learning data from databases with hierarchical structures, allowing for partial abstraction or detailing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If data items are extracted from databases with hierarchical structures without flexible abstraction control, then learning data can be generated, but the processing load increases and data efficiency decreases
Solution Approach 1:
The patent applies parameter changes by introducing a degree of abstraction parameter that can be adjusted for different data items. This allows the system to transform data from detailed representations to abstracted representations based on the required level of granularity, thereby optimizing processing efficiency and reducing unnecessary computational load while generating learning data.
Solution Approach 2:
The patent segments the data extraction process by applying different degrees of abstraction to different data items individually. Instead of uniformly processing all data at the same level of detail, the system divides the data items into groups requiring different abstraction levels, allowing efficient processing of each segment according to its specific requirements.
2Adaptability or versatility
If data items are uniformly abstracted or detailed, then processing is simplified, but the ability to tailor data to specific machine learning model needs is reduced
Solution Approach 1:
The patent applies local quality by allowing different data items to have different degrees of abstraction assigned to them individually. This enables the system to optimize each data item's representation according to its specific role in the machine learning model, rather than applying a uniform abstraction level across all data, thereby achieving high adaptability while managing complexity through targeted differentiation.
3Measurement precision
If all data items are kept at maximum detail, then data quality is high, but processing time and computational resources increase
Solution Approach 1:
The patent applies partial action by selectively applying detailed processing only to those data items that require high precision, while applying abstraction to other data items where full detail is unnecessary. This approach maintains high data quality for critical features while reducing overall processing time and computational resource consumption.
Data Source
AI summary
One preferred aspect of the present invention is a method of configuring a data model for learning data for machine learning, the method including, in a case where data items of a database as a basis of the learning data have a hierarchical structure of a degree of abstraction or a degree of detail, by using a filter that enables the degree of abstraction or the degree of detail of the data items to be designated for each of the data items and sorts the data items into an objective variable and an explanatory variable, configuring a data model that extracts a data item to be used for learning data among the data items from the database.