Risk Prediction Model with Efficient Feature Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current machine learning models require time-consuming feature engineering by subject matter experts, which may not comprehensively cover all features useful for a given problem, especially when dealing with high-dimensional raw data, leading to complex models and inefficient feature extraction.
Innovation Solution
A risk prediction model with efficient feature learning, comprising a feature learning model and a risk classification model, that extracts features from time-series data using a convolutional neural network and adjusts parameters to minimize a loss function, allowing for the removal of less influencing filters to reduce dimensionality and improve predictiveness.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If feature engineering is performed by subject matter experts to extract features from raw data, then the features can be useful for modeling, but the process is time consuming and may not comprehensively cover all features
Solution Approach 1:
The system enables automated feature learning where the machine learning model automatically learns and extracts features from raw data through neural network layers, eliminating the need for manual feature engineering by subject matter experts. The model performs self-service feature extraction that is both comprehensive and efficient.
Solution Approach 2:
The patent replaces the manual mechanical process of feature engineering with an automated computational system. Neural network layers automatically transform raw data into features through learned transformations, substituting human expert manual work with algorithmic feature learning.
2Measurement precision
If intricate structures with interconnected nodes are used in machine learning models to achieve high prediction accuracy, then prediction accuracy is improved, but model complexity increases
Solution Approach 1:
The patent extracts and removes less influential filters from the neural network model based on calculated influence scores. This pruning approach maintains the essential complex structure needed for high accuracy while eliminating redundant components that contribute to unnecessary model complexity.
Solution Approach 2:
The system dynamically adjusts model parameters by removing filters based on their influence scores. This parameter change optimizes the model structure, maintaining only the necessary complexity for accurate predictions while reducing overall model complexity through systematic parameter elimination.
3Loss of information
If high-dimensional raw data is processed to extract comprehensive features, then feature completeness is improved, but computational resources and processing time increase
Solution Approach 1:
The patent extracts only the most influential features from high-dimensional raw data by calculating influence scores for each filter. This selective extraction maintains feature completeness for important patterns while discarding redundant information, thereby reducing computational resource requirements.
Solution Approach 2:
The system applies different treatment to different features based on their local quality or importance. Influential filters are retained and processed in detail, while less influential filters are removed. This local quality approach ensures comprehensive feature extraction for important aspects while reducing computational burden for less critical features.
Data Source
AI summary
A method can be used to predict risk using machine learning models having efficient feature learning. A risk prediction model can be applied to time-series data associated with a target entity to generate a risk indicator. The risk prediction model can include a feature learning model for generating features from the time-series data. The risk prediction model can also include a risk classification model for generating the risk indicator. The feature learning model can include filters and can be trained. Parameters of the risk prediction model can be adjusted to minimize a loss function associated with risk indicators. An updated risk prediction model can be generated by removing a filter from an original set of filters based on influencing scores of the original filters. The risk indicator can be transmitted to a computing device for use in controlling access of the target entity to a computing environment.