Skip to content

init

segment_analysis ¤

Segment Analysis

Modular segment-based process analysis for production timeseries.

Detects time ranges from categorical signals (order numbers, part numbers) and applies them to process parameter UUIDs for per-segment analysis.

  • SegmentExtractor: Detect transitions in a categorical signal and extract time ranges.

  • extract_time_ranges: Detect value changes, return segment start/end/value/duration.

  • SegmentProcessor: Apply extracted time ranges to process data and compute profiles.

  • apply_ranges: Filter process data by time ranges, annotate with segment info.

  • compute_metric_profiles: Compute statistical metrics per UUID per segment.

  • ProfileComparison: Distance, clustering, similarity, and anomaly detection on profiles.

  • compute_distance_matrix: Pairwise distance matrix between groups.
  • cluster: Hierarchical clustering of items by metric similarity.
  • find_similar: Top-K most similar items to a target.
  • detect_anomalous: Flag items with unusual metric profiles.
  • detect_changes: Track metric shifts across consecutive segments per UUID.

  • find_similar_pairs: Find most similar (UUID, segment) pairs across all data.

  • TimeWindowedFeatureTable: Build ML-ready feature tables from segmented data.

  • compute_long: Metrics per (time_window, uuid, segment) in long format.

  • compute: Wide-format table with one row per time window, columns = {uuid}__{metric}.

  • FeaturePipeline: Generic pipeline builder for chaining any ts-shape class methods.

  • add_step: Register a @classmethod transform (DataFrame in → DataFrame out).
  • add_instance_step: Register an instance-based class step (DataHarmonizer, etc.).
  • add_lambda_step: Register a custom callable step.
  • run: Execute all steps and return the final DataFrame.
  • run_steps: Execute all steps and return intermediate results for debugging.