Skip to contents

Main anomaly detection pipeline for a single detector

pipe.Rd

This function processes one batch of time-series data for a single detector using a specified architecture and updates accumulated statistics over batches.

Usage

pipe(curr_batch, prev_batch, res.list, arch_params, verb = T)

Arguments

curr_batch

A `ts` object representing the current batch of data.

prev_batch

A `ts` object from the previous batch; used for boundary-aware processing.

res.list

A named list containing results so far, typically from `init_pipe()`.

arch_params

A named list or Rist containing architecture-related parameters:

arch

Main detection function for the pipeline.

n_missed

List with `Mh` and `Mt`, specifying how many points to include before and after.

DQ

Optional. DQ flag to include, e.g., `"BURST_CAT2"`.

P_update

Optional. Threshold for P0 filtering in update logic.

verb

Logical; if `TRUE`, prints progress messages.

Value

An updated `res.list` with fields:

proc

Processed dataframe for the current batch.

stat

Current batch statistics.

lamb

Lambda estimates (a, c) after update.

ustat

Updated cumulative statistics including `last_tcen`.

Details

The pipeline includes:

  • Concatenating previous and current batches

  • Running the detection algorithm (`arch`)

  • Filtering and aligning results with the current batch

  • Adding DQ flags and computing anomaly statistics

  • Computing significance probabilities (Poisson, Exponential, P0)

  • Updating cumulative statistics across batches

Examples

if (FALSE) { # \dontrun{
# Assume you have curr_batch and prev_batch as ts objects
dets <- c("H1", "L1")
arch_params <- config_pipe()
init <- init_pipe(dets = dets)
prev_batch <- init[[1]]
res.net <- init[[2]]

result <- pipe(curr_batch, prev_batch[["H1"]], res.net[["H1"]], arch_params)
} # }