Anomsmith Documentation
Anomsmith builds signal into judgment.
Anomaly detection lives between statistics and decisions. Most libraries stop at scores. Anomsmith carries work through scoring, thresholding, evaluation, and reporting with clear boundaries between each step.
Anomsmith treats anomaly detection as a workflow, not a function call. Data enters once. Methods produce scores. Tasks define intent. Workflows return evidence that supports action.
Anomsmith fits practitioners who work with operational time series. These series drift, spike, reset, and break assumptions. Anomsmith keeps those realities explicit instead of hiding them behind defaults.
Anomsmith does not replace forecasting or monitoring systems. It complements them. It shares typing and semantics with Timesmith so anomaly detection fits naturally into larger analytics systems.
Contents:
- Getting Started
- Architecture
- Network anomaly workflows
- Predictive maintenance platform
- Retiring
anomaly_detection_toolkit(asset_health) - Design Decisions
- Overview
- API Reference
ARIMADriftDetectorAlertAlertLevelAlertSystemBaseDetectorBaseScorerDashboardVisualizerEnsembleDetectorFailureClassifierFeatureExtractorIQRScorerIsolationForestDetectorLOFDetectorLSTMAutoencoderDetectorModelPerformanceTrackerPCADetectorPanelLikePredictiveMaintenanceSystemPyTorchAutoencoderDetectorRULEstimatorRealTimeIngestionRobustCovarianceDetectorRobustZScoreScorerScoreCombiningEnsembleDetectorSeasonalBaselineScorerSeriesLikeThresholdRuleVotingEnsembleVotingEnsembleDetectorWaveletDenoiserWaveletDetectorZScoreScoreradd_degradation_rates()add_rolling_statistics()aggregate_metrics_for_cloudwatch()apply_policy()assess_asset_health()assess_health_with_pca()backtest_detector()batch_predict()batch_score()calculate_confusion_matrix_metrics()calculate_lead_time()calculate_rul()classify_health_from_distance()compare_detectors()compare_survival_models()compute_concordance_index()compute_pca_health_thresholds()compute_performance_metrics()create_rul_labels()detect_anomalies()detect_concept_drift()discretize_rul()evaluate_detector()evaluate_policy()evaluate_survival_model()fit_survival_model_for_maintenance()plot_comparison_metrics()plot_pca_boundary()plot_reconstruction_error()plot_sensor_drift()predict_health_states_from_survival()predict_rul_from_survival()prepare_pm_features()rank_assets_by_risk()score_anomalies()sweep_thresholds()track_mahalanobis_distance()- Objects
- Primitives
- Workflows
- Platform (predictive maintenance)