Model Retraining & Recertification

Automate model lifecycle management with continuous validation

247

Active Models

12

Pending Retrain

5

In Training

28

Certified (30d)

45 d

Avg Cycle

8%

Avg Improvement
Model Retraining Queue
Credit Scoring Model
Model ID: MDL-001 Version: 2.3.1 → 2.4.0
Performance
In Training

Retraining Triggers

Accuracy Drop: -3.2%

Data Drift (PSI): 0.18

Time Since Last: 62 days

Training Progress

65% Complete ETA: 2 hours
Fraud Detection
Model ID: MDL-003 Version: 1.8.2 → 1.9.0
Scheduled
Pending

Retraining Triggers

Monthly Schedule: Due

False Positives: +2.1%

New Data Volume: 1.2M records

Training Progress

Scheduled: Tomorrow 2 AM Priority: Medium
Demand Forecast
Model ID: MDL-005 Version: 3.1.0 → 3.2.0
Drift
Completed

Retraining Triggers

Concept Drift: Detected

MAE Increase: +8.3%

Feature Drift: 3 features

Training Progress

Completed: 2 hours ago Duration: 4.5 hours
Recertification Pipeline
Model Version Performance Comparison
Model
Current Version
New Version
Accuracy Fairness Latency Certification Action
Credit Scoring 2.3.1 2.4.0
91.2% → 94.5%
+3.3%
0.08 → 0.04
-0.04
95 ms → 87 ms
-8 ms
Passed
Fraud Detection 1.8.2 1.9.0
88.7% → 89.2%
+0.5%
0.05 → 0.06
+0.01
120 ms → 115 ms
-5 ms
In Review
Demand Forecast 3.1.0 3.2.0
85.3% → 88.9%
+3.6%
0.03 → 0.02
-0.01
150 ms → 142 ms
-8 ms
Passed
Automated Retraining Triggers
Trigger Configuration
Trigger Type Metric Threshold Current Value Status Auto-Trigger
Performance Accuracy Drop > 2%
1.8%
Warning
Drift PSI Score > 0.2
0.15
Normal
Fairness Demographic Parity > 0.1
0.08
Warning
Time Days Since Training > 60
45
Normal
Volume New Data Records > 1M
850K
Warning
Trigger Statistics

Total Triggers (30d)

89

Auto-Triggered

67

Manual Override

22

Retraining History & Performance
Retraining Frequency (6 Months)
02040 JulAugSepOctNovDec
Models Retrained
Performance Improvement Trend
020 JulAugSepOctNovDec
Avg Improvement %
Recent Certification Records

Customer Churn v@cert.Version

Certified

All governance checks passed, performance improved by 5.2%

Certifier: Sarah Chen | Duration: 3.5 hours
Today

Risk Assessment v@cert.Version

Conditional

Certified with monitoring requirements for fairness metrics

Certifier: Mike Johnson | Duration: 5.2 hours
Yesterday

Price Optimizer v@cert.Version

Failed

Failed fairness requirements, requires additional bias mitigation

Certifier: Lisa Wang | Duration: 2.8 hours
3 days ago
An unhandled error has occurred. Reload 🗙