Growth Potential Estimation via Time-Series Graph Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing growth potential estimation systems fail to accurately capture the features of time-series changes in company activities, leading to decreased estimation accuracy, as they do not sufficiently grasp the complex interrelations between transaction and attribute changes in company activities.
Innovation Solution
A growth potential estimation system that utilizes an estimation model representing the relations between transaction information, account time-series information, and company attribute information to estimate future growth potential by extracting and analyzing features from a graph structure that changes over time, employing algorithms like TGFN, STAR, and Netwalk to determine explanatory variables and their importance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional regression analysis and score calculation methods are used to estimate growth potential, then the system is simple to operate and implement, but the estimation accuracy is insufficient because the system cannot sufficiently grasp the features of time-series changes in company activities
Solution Approach 1:
The patent segments the estimation process into multiple components: extracting time-series data from multiple sources (transaction information, account information, attribute information), analyzing temporal patterns separately, and then integrating these analyses through machine learning models. This segmentation allows the system to capture complex time-series features while maintaining manageable system architecture.
Solution Approach 2:
The patent introduces machine learning models as intermediary components between raw data and estimation results. These models (including deep learning architectures) act as mediators that automatically extract temporal features and patterns from multi-source data, transforming complex time-series information into meaningful growth potential estimates without requiring manual feature engineering.
2Measurement precision
If the system analyzes multiple types of time-series information (transaction, account, attribute) to improve estimation accuracy, then the measurement precision improves, but the device complexity increases due to the need to process and integrate multiple data sources
Solution Approach 1:
The patent employs universal machine learning frameworks that can process multiple data types (transaction, account, attribute information) through the same architectural structure. The deep learning models are designed to handle heterogeneous data sources uniformly, extracting temporal features from each source type while maintaining a consistent processing pipeline, thereby reducing the complexity increase that would result from separate processing systems for each data type.
Data Source
AI summary
A growth potential estimation system 30 is provided with: an estimation model 31 that represents a relationship between transaction information 310 (representing a time-series change of company-to-company transaction relations of an intended company), account time-series information 313 (representing a time-series change of deposits and withdrawals of accounts of the intended company), and intended company attribute information 314 (representing a time-series change of the attribute of the intended company) of the intended company for a first period and the growth potential 315 of the intended company after the first period; and an estimation unit 32 for estimating the growth potential of the intended company after a second period on the basis of transaction information 300, account time-series information 303, and company attribute information 304 for a second period that is later than the first period.