Multi-Subset Feature Prediction for Abnormal Instance Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods fail to accurately detect abnormal instances when multiple features become abnormal, as they rely solely on prediction models using all features except the target variable as explanatory variables, leading to incorrect determination of normality.
Innovation Solution
An information processing apparatus and method that acquires an instance with multiple features, uses subsets of features excluding the target feature as explanatory variables, and outputs prediction results from multiple prediction models to determine the degree of abnormality, allowing for more accurate detection even when multiple features are abnormal.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a prediction model uses all features except the target variable as explanatory variables, then the prediction accuracy for single-feature abnormalities is improved, but the detection accuracy for multi-feature abnormalities deteriorates
Solution Approach 1:
The patent segments the feature set into multiple subsets, where each subset excludes a different target feature. Instead of using all features except one target variable, the method creates multiple prediction models, each trained on a different subset of features. This segmentation allows the system to evaluate abnormalities from multiple perspectives, preventing the masking effect where multiple abnormal features cancel each other out in a single comprehensive model.
Solution Approach 2:
The patent applies local quality by creating prediction models with different local feature compositions. Each prediction model is trained on a specific subset of features tailored to predict a particular target feature, rather than using a uniform feature set for all predictions. This localized approach ensures that each model evaluates abnormalities in its specific context, improving the detection of multi-feature abnormalities.
2Reliability
If multiple prediction models with different feature subsets are used, then the detection accuracy for multi-feature abnormalities is improved, but the computational complexity increases
Solution Approach 1:
The patent divides the computational task into multiple smaller prediction models, each handling a specific feature subset. This segmentation reduces the computational burden on individual models compared to a single comprehensive model, while the parallel structure allows for efficient implementation. Each model is trained on fewer features, reducing training complexity and inference time.
Solution Approach 2:
The patent uses partial feature sets for each prediction model rather than all features. By training each model on a subset of features (excessive action in terms of model数量 but partial in feature usage), the system achieves better detection accuracy without requiring each individual model to process the entire feature space, thus balancing computational complexity with detection reliability.
Data Source
AI summary
In order to achieve an object to make it possible to more accurately detect an abnormal instance, an information processing apparatus includes: an acquisition means (21) that acquires an instance expressed as a set of a plurality of features; a prediction means (22) that outputs a plurality of prediction results which are obtained by using (i) as a target variable, at least one of the plurality of features which are included in the instance and (ii) as explanatory variables, a plurality of subsets of features obtained by excluding the at least one feature from the plurality of features, the subsets being different from each other; and an abnormality degree output means (23) that outputs a degree of abnormality of the instance, with reference to the plurality of prediction results.