CAN Signal Predictive Modeling With Correlation-Based Feature Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for AI-based predictive modeling of controller area network (CAN) signals require manual configuration and design, which is time-consuming.
Innovation Solution
A method and apparatus for predictive modeling of CAN signals using AI, involving training data management that includes analyzing CAN signals, selecting features based on correlation, removing multicollinearity, data augmentation, and utilizing pre-trained models with output layers.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If manual configuration and design of AI model architecture is performed, then model customization and adaptability are improved, but development time and complexity increase
Solution Approach 1:
The patent applies preliminary action by using pre-trained models that have already been trained on large datasets before being deployed for CAN signal prediction. This allows the system to start with a model that already possesses general predictive capabilities, significantly reducing the time required for model development while maintaining adaptability through fine-tuning on specific CAN datasets.
Solution Approach 2:
The patent employs copying by utilizing pre-trained models that have been developed and trained independently. These pre-trained models serve as templates or copies that can be directly applied to CAN signal prediction tasks, eliminating the need to build models from scratch and reducing development time while preserving adaptability through task-specific adjustments.
2Measurement precision
If extensive data preprocessing and feature selection is performed, then model accuracy is improved, but processing complexity increases
Solution Approach 1:
The patent applies the taking out principle by selectively extracting only the most relevant features from CAN signals using correlation-based feature selection. Instead of processing all available data, the system identifies and extracts key features that have high correlation with target variables, thereby improving prediction accuracy while reducing processing complexity.
Solution Approach 2:
The patent applies local quality by applying different preprocessing strategies to different parts of the data processing pipeline. Specifically, correlation-based feature selection is applied to numerical features, while other preprocessing techniques are applied to categorical features, allowing each type of data to be processed with the most appropriate method and reducing overall complexity.
3Reliability
If data augmentation techniques are applied, then model robustness is improved, but training time increases
Solution Approach 1:
The patent applies partial action by implementing data augmentation only for specific cases where it is most beneficial. The system uses SMOTE (Synthetic Minority Over-sampling Technique) to generate synthetic samples primarily for underrepresented classes or scenarios, rather than applying aggressive augmentation to all data. This selective approach improves model robustness for critical cases while minimizing the increase in training time.
Data Source
AI summary
A method for predictive modeling of a controller area network (CAN) signal includes managing training data based on a CAN signal. Managing the training data includes analyzing the CAN signal and selecting a feature based on a degree of correlation. The method also includes performing training on an artificial intelligence (AI) model based on the training data.