Multioutput-Multilabel ML Performance Metrics Computation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current machine learning libraries do not support the computation of performance metrics for multioutput-multilabel machine learning models, which are essential for tracking performance and making informed decisions.
Innovation Solution
The proposed solution involves a method to generate performance metrics for multioutput-multilabel machine learning models by using micro and macro values. These values are calculated based on aggregated and averaged true negative, true positive, false negative, and false positive values, respectively.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If traditional machine learning libraries are used, then binary classification metrics can be computed, but multioutput-multilabel model performance cannot be measured
Solution Approach 1:
The patent segments the computation of performance metrics by introducing separate micro-level and macro-level calculation approaches. Micro metrics compute performance for each individual label and output independently, then aggregate results. Macro metrics compute performance across all labels and outputs collectively. This segmentation enables comprehensive measurement for multioutput-multilabel models while maintaining compatibility with traditional binary classification frameworks.
2Measurement precision
If micro and macro values are computed separately, then comprehensive performance metrics are obtained, but computational complexity increases
Solution Approach 1:
The patent merges micro-level and macro-level metric computations into a unified framework. Both micro and macro metrics are calculated using the same underlying confusion matrix structure, allowing comprehensive performance assessment without requiring entirely separate computational systems. The confusion matrix serves as a common foundation that integrates both approaches efficiently.
3Loss of information
If performance metrics are computed for each class and output individually, then detailed performance insights are obtained, but data processing requirements increase
Solution Approach 1:
The patent extracts essential performance information by computing metrics at two distinct levels: micro-level extraction provides detailed per-class and per-output performance, while macro-level extraction provides overall model performance. This extraction approach retrieves only the necessary metric values from the confusion matrix without requiring storage or processing of all intermediate prediction details, thus retaining performance information while managing data volume efficiently.
Data Source
AI summary
The present disclosure relates to machine learning (ML) models, and more particularly to novel techniques for computing performance metrics for Multioutput-Multilabel ML models. Novel techniques are described for computing the performance metrics in a parallel and distributed without having to store the entire dataset for which metrics are to be computed in the memory of a data processing system. Novel data structures are provided for performing the computations.