Automated Feature Engineering Using Negative Inferences
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional automated feature engineering in machine learning lacks the capability to incorporate complex logical inferences, while manual feature engineering is time-consuming and impractical for large datasets, limiting the effectiveness of machine learning models in applications like treatment selection.
Innovation Solution
A method that uses co-occurrence rates to identify inverse relationships between concepts, enabling the generation of test data with negative inferences for training machine learning models, thereby enhancing the quality of feature engineering and model accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If automated feature engineering is used, then processing speed and scalability are improved, but the capability to incorporate complex logical inferences deteriorates
Solution Approach 1:
The patent introduces an intermediary module that automatically generates logical inference rules (including negative inferences) from training data, which then serve as additional features for the machine learning model. This intermediary process bridges the gap between automated processing and complex logical reasoning by generating structured inference rules that the model can utilize without requiring manual engineering of each logical relationship.
2Reliability
If manual feature engineering is used to include complex logical inferences, then the quality of feature engineering is improved, but time consumption and cost increase
Solution Approach 1:
The system performs self-service by automatically analyzing training data to identify and generate logical inference rules without requiring manual intervention. The automated feature engineering system serves itself by producing the complex logical features that would otherwise require manual engineering, thereby maintaining high feature quality while eliminating time consumption and cost associated with manual processes.
Solution Approach 2:
The patent performs preliminary action by pre-computing logical inference rules from the training data before the main machine learning training process. This preliminary feature engineering step prepares the data in advance, allowing the model to benefit from complex logical relationships without requiring time-consuming manual feature engineering during deployment or model development.
3Adaptability or versatility
If the number of potential inferences across all attributes is increased, then the comprehensiveness of feature engineering is improved, but system complexity and computational cost increase
Solution Approach 1:
The patent applies local quality by generating logical inference rules specifically targeted to the relationships present in the given training data, rather than attempting to pre-compute all possible inferences across all potential attributes. The system focuses computational resources on deriving relevant inference rules locally from the actual data patterns, thereby achieving comprehensiveness for the specific application without overwhelming system complexity.
Solution Approach 2:
The system dynamically adjusts the parameters of feature engineering based on the characteristics of the training data. Instead of using a fixed comprehensive set of inference rules, the patent changes the parameters of which inferences to generate based on data analysis, allowing the system to achieve appropriate comprehensiveness while controlling complexity by adapting to the specific requirements of each dataset.
Data Source
AI summary
Method and apparatus for performing feature engineering using negative inferences are provided. One example method generally includes identifying a plurality of concepts and analyzing a corpus of documents to determine a first co-occurrence rate for a first concept and a second concept in the plurality of concepts. The method further includes analyzing the corpus of documents to determine a second co-occurrence rate for the second concept and at least a third concept of a set of concepts related to the first concept and determining an inverse relationship between the second concept and the third concept. The method further includes generating test data for training a machine learning model including a negative inference between the second concept and the third concept and training the machine learning model using the test data.