Ensemble Learning TRIZ Cases
- Edge Classification Accuracy via Auto-Feature Discriminator Offset Adjustment
Auto-feature discriminator adjusts classification thresholds using local data to improve remote livestock monitoring accuracy.
- Dictionary-Based Deduplication for ML Malware Training Sets
Vector binarization and dictionary deduplication remove redundant malware samples, reducing training data volume by 30% while maintaining detection coverage.
- Annotator Performance Assessment via Baseline Model Comparison
Trains annotator-specific models against a baseline to score performance and identify inaccuracies for targeted retraining.
- Machine Learning Fraud Detection via Activity Normalization
Normalizing diverse activity data enables machine learning models to detect subtle first-party fraud early, reducing financial losses.
- AI Model Training Data Refinement via Synthetic Sampling
A machine learning system applies under-sampling and over-sampling techniques to refine training data sets before model configuration.
- Expert Diversification via Bidding Mechanism in Ensemble Learning
Bidding mechanism allocates training data to experts based on performance, resolving ineffective diversification and improving decision accuracy.
- Distributed AI Ensemble Fusion for Heterogeneous Node Training
Segments centralized training across independent nodes using ensemble fusion to balance class distribution and reduce data transmission costs.
- ML-Based Synchronization Anomaly Detection in Heterogeneous RAN Nodes
Automated ML analysis of alarm logs identifies root causes of timing errors, eliminating manual expert intervention for heterogeneous RAN nodes.