Skip to contents

Path anomalies (HYPA) in sequences

Source: R/dispatch.R
hypa.default.Rd

The sequence method of hypa(). Builds the De Bruijn graph of order order from the sequences and scores every path of order + 1 states against a hypergeometric null whose propensities are fitted to the observed path counts: paths observed far more often than the null predicts are over-represented, far less often under-represented.

Usage

# Default S3 method
hypa(
  x,
  order = 2L,
  alpha = 0.05,
  min_count = 5L,
  p_adjust = "BH",
  type = c("all", "over", "under"),
  order_by = c("sig", "ratio", "freq", "path"),
  n = 10L,
  action = NULL,
  actor = NULL,
  time = NULL,
  session = NULL,
  time_threshold = 900,
  timezone = "UTC",
  ...
)

# S3 method for class 'hypernets_hypa'
print(x, n = NULL, ...)

Arguments

x

Sequences in any form described in sequence-input: a long event table (with action), a wide data.frame, a list of vectors, or a model object carrying its sequences.

order

Integer (or integer vector) of De Bruijn orders. Default 2.

alpha

Significance level of the anomaly labels, in (0, 0.5). Default 0.05.

min_count

Integer. Paths observed fewer times are always labelled "normal". Default 5.

p_adjust

Multiplicity correction: any stats::p.adjust method or "none". Default "BH"; the two tails are adjusted separately.

type

Which anomalous paths the printed result lists: "all" (default), "over" (paths observed more often than expected) or "under" (less often).

order_by

How the printed paths are ordered: "sig" (default, smallest p-value of the path's own direction first), "ratio" (largest deviation first: highest observed/expected ratio for over-represented paths, lowest for under-represented ones), "freq" (most observed first) or "path" (alphabetical).

n

Number of paths printed per direction. Default 10.

action, actor, time, session, time_threshold, timezone

Long-format arguments (see sequence-input); leave the column names NULL for wide, list or model input.

...

Must be empty: an argument that only the hypergraph method takes (top) raises hypernets_bad_input.

Value

A net_hypa object (also a cograph_network) that prints the selected anomalous paths. type, order_by and n change only what is printed and the default of hg_get(); every path is scored. Read the tables with hg_get(): the anomalous paths (what = "anomalies", the default), every scored path ("scores"), the over- or under-represented ones ("over", "under"), or the anomalous paths as pathways ("pathways").

References

LaRock, T., Nanumyan, V., Scholtes, I., Casiraghi, G., Eliassi-Rad, T., and Schweitzer, F. (2020). HYPA: Efficient detection of path anomalies in time series data on networks. Proceedings of the 2020 SIAM International Conference on Data Mining, 460-468. doi:10.1137/1.9781611976236.52

See also

hypa.net_hg() for the same null on hypergraph co-occurrence.

Examples

seqs <- list(c("a", "b", "c", "a", "b", "c"), c("x", "b", "d", "x", "b", "d"),
             c("a", "b", "c", "a", "b", "c"), c("x", "b", "d", "x", "b", "d"))
fit <- hypa(seqs, order = 2, min_count = 1, type = "over")
fit
#> Path anomalies (HYPA, order 2): 0 of 6 paths anomalous at alpha = 0.05 (0 over, 0 under, BH)
#> (no rows)
hg_get(fit)
#> [1] path      count     expected  ratio     p_adj     direction
#> <0 rows> (or 0-length row.names)