Feature Value Generation Device Using Similarity Functions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing methods for generating feature candidates in data mining are labor-intensive, particularly when combining multiple information sources, as they require complex processing and substantial analyst effort.
Innovation Solution
A feature generating device and method that acquire tables with prediction targets, use similarity functions to combine records from these tables based on shared attributes, and apply reduction conditions to reduce data, thereby simplifying the feature generation process and reducing analyst labor.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If complex processing methods are used to generate feature candidates by combining multiple tables, then the quality and variety of feature candidates improve, but the labor required by analysts increases substantially
Solution Approach 1:
The system enables self-service feature generation by automatically combining multiple tables using similarity functions and reduction methods without requiring substantial analyst intervention. The computer executes the feature generation process autonomously based on predetermined conditions, allowing analysts to obtain high-quality feature candidates with minimal manual effort.
Solution Approach 2:
The patent replaces manual mechanical processing by analysts with automated computer-based processing. The system uses algorithmic similarity functions and automated reduction methods to perform table combinations, substituting the mechanical labor of analysts with automated computational processes that maintain high feature quality while reducing human effort.
2Measurement precision
If multiple map conditions are considered for different analysis targets, then the accuracy of feature matching improves, but the processing complexity increases
Solution Approach 1:
The system manages processing complexity by dynamically changing parameters such as similarity thresholds and reduction methods based on the analysis target. The computer automatically adjusts these parameters to maintain high matching accuracy while optimizing the processing complexity for each specific analysis scenario, preventing the system from becoming overly complex.
Solution Approach 2:
The patent introduces dynamic adaptability by allowing the system to adjust map conditions and similarity functions based on different analysis targets. The processing complexity is managed dynamically rather than being fixed, enabling the system to maintain high accuracy while adapting to different scenarios without requiring complex processing for all cases simultaneously.
3Quantity of substance
If comprehensive table combination processing is performed, then more feature candidates are generated, but the time required for processing increases
Solution Approach 1:
The system applies partial action by using reduction methods that selectively process only the necessary portions of combined tables. Instead of processing all possible combinations exhaustively, the computer applies reduction conditions to generate sufficient feature candidates without performing excessive processing, thereby reducing the time required while maintaining an adequate number of feature candidates.
Solution Approach 2:
The patent employs preliminary action by applying reduction methods and similarity conditions before full feature generation. The computer pre-processes the combined tables using predetermined reduction rules, which prepares the data in advance and enables faster subsequent processing, thus generating comprehensive feature candidates more quickly.
Data Source
AI summary
A table acquiring means 381 acquires a first table including prediction objects and first attributes, and a second table including second attributes. A receiving means 382 receives a similarity function and condition for similarity used to calculate the similarity between the first attribute and the second attribute. A feature generating means 383 generates feature candidates able to affect a prediction object using a combination condition for combining a record in the first table including the value of a first attribute satisfying the condition with a record in the second table including the value of a second attribute satisfying the similarity calculated with the value of the first attribute and the value of the second attribute using the similarity function, and using a reduction method for a plurality of records in the second table and a reduction condition represented by the column to be aggregated. A feature selecting means 384 selects an optimum feature for the prediction from the feature candidates.