Semantic Target Prediction Using Causal Graphs and Mixed Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing futurecasting models struggle with accurately predicting targets using a combination of structured and unstructured data, particularly in mid-to-long-term forecasts, and lack clarity in presenting the basis for predictions.
Innovation Solution
A target prediction method and system that analyzes relationship information between a target and its influence variables at a semantic level, filtering and integrating structured and unstructured data to generate a basis for future outlooks, using named entity recognition, sentiment analysis, and causal relationship graphs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If futurecasting models utilize both structured and unstructured data for predictions, then the comprehensiveness of data utilization is improved, but the accuracy of numerical value predictions deteriorates
Solution Approach 1:
The patent segments the data processing into distinct pipelines: one for structured data (numerical values) and another for unstructured data (text analysis). Each pipeline processes its specific data type through dedicated methods, with structured data going through feature extraction and unstructured data going through NLP processing, before their results are integrated for final prediction.
Solution Approach 2:
The patent introduces an intermediary processing layer that converts unstructured text data into structured feature representations. This intermediary layer processes text through NLP methods (entity recognition, sentiment analysis, topic modeling) and transforms the output into a format that can be combined with existing structured numerical data, enabling unified processing while maintaining accuracy.
2Duration of action of moving object
If futurecasting models focus on mid-to-long-term predictions, then the strategic value is improved, but the prediction reliability deteriorates
Solution Approach 1:
The patent performs preliminary actions by establishing causal relationship graphs and identifying key influence variables before making predictions. The system pre-processes data to understand causal structures and relationships, creating a foundation that guides subsequent predictions and allows for more reliable mid-to-long-term forecasts by accounting for causal mechanisms rather than just pattern recognition.
Solution Approach 2:
The patent incorporates feedback mechanisms where prediction results are used to refine the causal relationship graphs and influence variable identifications. The system continuously learns from prediction outcomes and adjusts its causal models, improving reliability over time for mid-to-long-term forecasts by adapting to changing conditions and validating causal assumptions against actual outcomes.
3Loss of information
If futurecasting models provide detailed prediction bases, then the explainability is improved, but the computational complexity increases
Solution Approach 1:
The patent extracts and highlights only the most relevant causal relationships and influence variables from the complex data processing. By identifying and presenting only the key causal pathways and their impacts, the system provides detailed explainability for predictions while avoiding the computational burden of processing and presenting all intermediate processing steps and data transformations.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A target prediction method for predicting a future outlook of a target performed by a computing device according to an embodiment of the present disclosure is configured to collect related structured and unstructured data when a user requests target prediction, analyze the relationship between the target and a variable affecting the target at a semantic level, and then compute a target outlook of a future.