ARIMA upper limit alerter

arimaUpperLimitAlert

#!/usr/bin/env r
#######################################################################################################################
#    
#    This script was inspired by this post:
#        http://stats.stackexchange.com/questions/152644/what-algorithm-should-i-use-to-detect-anomalies-on-time-series
#
#    It creates a timeseries model for all the previous periods except the most recent one, calculates confidence
#    intervals for alerts amd warnings, and then returns:
#        2: most recent value falls outside the alert upper limit
#        1: most recent value falls outside the warning upper limit
#        0: most revent value is within both upper limits
#    
#######################################################################################################################

suppressWarnings(suppressPackageStartupMessages(library(RMySQL)))
suppressWarnings(suppressPackageStartupMessages(library(dplyr)))
suppressWarnings(suppressPackageStartupMessages(library(zoo)))
suppressWarnings(suppressPackageStartupMessages(library(forecast)))

con <- dbConnect(MySQL(), host = "fqdn.of.host", user = "myuser", pass = "Mr. Creosote", dbname = "metrics")

sql <- "select timestamp, someMetric from [someTable] order by timestamp"

suppressWarnings(data <- dbGetQuery(con, sql))

data <- data[, c('timestamp', 'someMetric')]

actual <- tail(data, n = 1)$someMetric

dbDisconnect(con)

data <- head(data, nrow(data) - 1)

data <- data[, c('timestamp', 'someMetric')]

data.zoo <- zoo(data$someMetric, order.by=data$timestamp)

data.model <- auto.arima(data.zoo, ic = c("bic"))

data.prediction.warningLimits <- forecast(data.model, h=1, level=0.99)
data.prediction.alertLimits <- forecast(data.model, h=1, level=0.999)

ifelse(
  actual < as.numeric(data.prediction.warningLimits$upper),
  resultCode <- 0,
  ifelse(
    actual < as.numeric(data.prediction.alertLimits$upper),
    resultCode <- 1,
    resultCode <- 2
  )
)

cat(resultCode)
cat("\n")