Graph-Based Relevance Calculation for Small Data Sets
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional techniques struggle to accurately determine the relevance between a target category and another category, especially when the number of data points in the target category is small, and often fail to consider the evaluation index features, leading to decreased calculation accuracy.
Innovation Solution
An information processing apparatus that calculates the degree of relevance between a target category and another category by using a graph structure representing relational data, where a score is calculated based on the degree of relationship between data pairs and an evaluation index, enabling accurate relevance determination even with small data sets and varied evaluation indexes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional clustering or grouping techniques are used to extract another category from relational data, then the process is simple to implement, but the calculation accuracy of the degree of relevance decreases when the number of data points in the target category is small or when evaluation index features are not considered
Solution Approach 1:
The patent segments the relevance calculation process into multiple components: graph construction from relational data, score calculation for each data pair based on relationship degree, and evaluation index-based relevance determination. This segmentation allows each component to be optimized independently, improving overall accuracy without overwhelming complexity
Solution Approach 2:
The patent introduces a graph structure dimension to represent relational data, transforming traditional flat data relationships into a multi-dimensional graph representation. This enables capturing complex relationships between data pairs that conventional clustering methods miss, particularly when target category data is scarce
2Measurement precision
If conventional techniques perform clustering from purchase data using non-negative matrix decomposition or define unique indexes based on purchase trends, then these methods work for large data sets, but they fail to provide accurate relevance when the number of pieces of data in the target category is small
Solution Approach 1:
The patent introduces an intermediary evaluation index that mediates between the graph structure and relevance calculation. This evaluation index aggregates information from multiple data pairs and their relationships, enabling accurate relevance determination even when individual target category data points are few
Solution Approach 2:
The patent replaces conventional matrix decomposition mechanical operations with a graph-based score calculation system. This substitution allows flexible incorporation of evaluation indexes and relationship degrees, adapting to small data sets where matrix decomposition fails
Data Source
AI summary
An information processing apparatus 10 includes a processing unit 20. The processing unit 20 calculates, for each pair of pieces of data defined in a graph representing relational data 16A indicating a relationship between a plurality of pieces of data in a graph structure, a degree of relevance between a target category to be analyzed and another category other than the target category on the basis of a score calculated according to a degree of relationship between a pair of pieces of data constituting the pair and an evaluation index for evaluating the degree of relevance between the categories.