library("knitr")
opts_chunk$set(cache=TRUE)
library("tidyverse")
library("lubridate")
library("detectR")

First we clean up the piwick logs. Two variables allow to filter the dataset more effectively:

logs_board <- get_aggregated_piwick(system.file("extdata/complete_log_clean.csv", package = "detectR"), absolute_days=20, relative_days=0.9)

Then we get the main variable of time variability, which includes seasonal (weekly) variation, general trend on revues.org.

times_decomposition_board <- get_calendar_time_series(logs_board)

We compensate the recorded view with the time variability variables.

logs_board$aggregated_frequent <- correct_time_series(logs_board$aggregated_frequent, times_decomposition_board)

We proceed with anomaly detection, using a restricted cubic spline with 3 knots.

anomaly_detection <- detection_anomalies_rcs(logs_board$aggregated_frequent)

We can get a graph of the anomalies for one article

graph_anomalies_rcs(anomaly_detection, "mots.revues.org/21665")

It is also possible to get the same with adjusted size for sigma ratio for each anomaly (ie to what extent does it affect the variable)

graph_anomalies_rcs(anomaly_detection, "asiecentrale.revues.org/488", show_sigma = TRUE)


joelgombin/detectR documentation built on May 21, 2019, 7:55 a.m.