Model Variable Candidate Generation for Prediction Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In big data analysis, the increasing number and variability of data types and items make it difficult to select relevant explanatory variables for model generation, leading to decreased prediction accuracy and delayed model updates due to reliance on human operation and lack of established methods for generating valid candidates.
Innovation Solution
A model variable candidate generation device that includes a data input unit, item property preliminary setting unit, data property determination unit, and variable candidate generation unit, which sets and determines properties of items and data to generate explanatory variable candidates based on item and data properties, reducing manual operations and improving prediction model accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If human operation is used to select explanatory variables, then prediction model accuracy can be maintained through expert judgment, but the process becomes time-consuming and difficult to scale with increasing data volume
Solution Approach 1:
The system performs preliminary classification of data into multiple categories based on data properties before variable selection. This preliminary action organizes the data structure in advance, enabling automated algorithms to efficiently identify candidate variables without manual intervention, thus reducing time loss while maintaining accuracy through systematic preprocessing
Solution Approach 2:
The patent introduces an automated variable selection apparatus as an intermediary between raw data and prediction model generation. This intermediary systematically classifies data properties, identifies candidate variables through automated algorithms, and selects explanatory variables based on statistical criteria, replacing manual human operation while preserving accuracy through structured processing
2Adaptability or versatility
If all data items are considered as explanatory variable candidates, then comprehensive model coverage is achieved, but the complexity of variable selection increases dramatically
Solution Approach 1:
The patent segments the vast set of data items into multiple categories based on data properties (e.g., numerical, categorical, temporal). This segmentation divides the complex selection task into manageable subsets, reducing selection complexity while maintaining comprehensive coverage by ensuring each segment is systematically evaluated for its relevance as an explanatory variable candidate
Solution Approach 2:
The system applies different selection criteria and analysis methods to different data categories based on their specific properties. Instead of treating all data items uniformly, it tailors the variable selection approach to each data type's characteristics, reducing overall complexity by handling each category appropriately while maintaining comprehensive model coverage through localized optimization
3Productivity
If automated variable selection is implemented, then processing efficiency improves, but the ability to handle diverse and varying data types becomes more challenging
Solution Approach 1:
The patent implements a universal data classification framework that can handle multiple data types (numerical, categorical, temporal, textual) through a single automated apparatus. The system uses property-based classification that adapts to various data formats, enabling one unified system to process diverse data types efficiently while maintaining high productivity through automated standardized procedures
Data Source
AI summary
A model variable candidate generation device generating explanatory variable candidates to be used as candidates for an explanatory variable in generation of a prediction model includes: a data input unit inputting analysis data each entry having one or more items and the items having item values; a first item determination unit preliminarily setting properties of the items included in the analysis data as first item properties; a data property determination unit determining data properties being of the analysis data on the basis of the first item properties; a second item determination unit determining properties of the items included in the analysis data as second item properties on the basis of the data properties of the analysis data; and a variable candidate generation unit generating the explanatory variable candidates by selecting from the items or processing the items on the basis of the second item properties.