Target Prediction Using Causal Fusion of Structured and Text Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing future-casting models face limitations in accurately predicting targets due to the challenge of integrating structured and unstructured data, particularly in predicting numerical values, and lack clarity in causal relationships between targets and influence variables.
Innovation Solution
A target prediction method and system that utilizes a language model to detect target influence variables at a semantic level, filtering structured and unstructured data, and generating relationship information to predict short-term to long-term outlooks based on causal relationships.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If future-casting models use large-scale language models to predict future trends, then the ability to process natural language and text data is improved, but the accuracy in predicting numerical values by simultaneously utilizing structured and unstructured data deteriorates
Solution Approach 1:
The patent segments the prediction system into multiple specialized components: a language model processing unit for unstructured text data, a structured data processing unit for numerical data, and an integration unit that combines their outputs. This segmentation allows each component to optimize for its specific data type while maintaining overall prediction accuracy through coordinated integration.
Solution Approach 2:
The patent introduces an intermediary integration mechanism that bridges the language model and structured data processing. This intermediary component transforms and aligns the outputs from both data sources, enabling their effective combination without compromising the precision of numerical predictions while preserving the natural language processing capabilities.
2Reliability
If the system integrates multiple data sources and processing steps to improve prediction accuracy, then the prediction reliability is improved, but the system complexity increases
Solution Approach 1:
The patent implements a multi-functional integration unit that performs multiple operations: data alignment, feature extraction, fusion, and output generation. This universal component handles various data types and processing requirements through a single coordinated system, reducing overall system complexity while maintaining high prediction reliability through comprehensive data integration.
3Loss of information
If the model processes both structured and unstructured data simultaneously to provide comprehensive predictions, then the information completeness is improved, but the difficulty of detecting and measuring relationships between variables increases
Solution Approach 1:
The patent segments the relationship detection process into distinct phases: relationship identification in unstructured text, relationship extraction from structured data, and relationship integration. Each phase focuses on specific types of relationships, making detection and measurement more manageable while preserving complete information from both data sources.
Solution Approach 2:
The patent introduces intermediary processing layers that transform raw data from both structured and unstructured sources into standardized relationship representations. These intermediaries facilitate the detection and measurement of relationships by providing a common framework for integrating diverse data types, reducing the complexity of analyzing variable relationships across multiple sources.
Data Source
AI summary
A predictive generation method for predicting a future outlook of a target performed by a computing device or a processor may collect related structured and unstructured data when a user requests predictive generation, analyze the relationship between the target and a variable affecting the target at a semantic level, and compute a target outlook of a future.