Module 16
Model Registry & Governance
Real trained ML models (logistic regression, DBSCAN) plus transparent engineering heuristics.
coverage-need-logreg
Logistic Regression (trained live, in-browser, pure TS) · owner RF Planning
trained livedeployed
Features: rsrp_dbm, sinr_db, indoor, throughput_gap, demand_mbps, latency_ms, mobility_vehicular
Dataset: Built-in Fifth Settlement · 321 sites · 420 users
Validation: {"accuracy":0.952,"precision":0.986,"recall":0.959,"f1":0.973,"auc":0.984,"logLoss":0.1567,"trainMs":0}
weak-zone-dbscan
DBSCAN (unsupervised, Ester et al. 1996, pure TS) · owner RF Planning
deployed
Features: lat, lon, eps=180m, minPts=4
Dataset: Built-in Fifth Settlement · 321 sites · 420 users
Validation: {"algorithm":"density-based","distance":"haversine"}
coverage-need-rf
Random Forest (embedded) · owner RF Planning
deployed
Features: rsrp, sinr, indoor, throughput_gap, demand
Dataset: Built-in Fifth Settlement · 321 sites · 420 users
Validation: {"auc":0.91,"accuracy":0.87}
capacity-demand-gbm
Gradient Boosted Trees · owner Capacity
deployed
Features: active_users, video_ratio, vr_ratio, current_capacity, latency
Dataset: Built-in Fifth Settlement · 321 sites · 420 users
Validation: {"mae":62,"r2":0.83}
⚠ Underestimates VR-heavy weekends
security-risk-mlp
MLP + rule ensemble · owner Security
deployed
Features: signaling_rate, handover_freq, auth_failures, beam_switches
Dataset: Built-in Fifth Settlement · 321 sites · 420 users
Validation: {"precision":0.88,"recall":0.79}
band-selector-heuristic
Transparent heuristic · owner RF Planning
deployed
Features: indoor, demand, sinr
Dataset: Built-in Fifth Settlement · 321 sites · 420 users
Validation: n/a (heuristic)
⚠ Not a trained model — rule-based.
Take action
Not signed in — actions apply locally to this session only.