R의 ARIMA 모형의 예제입니다.
서버의 메모리의 사용량의 추이를 보고 얼마 후에 고갈되는지를 예측하는 코드입니다.
물론 예측력은 많이 떨어지고 현실성이 없을 수 있습니다.
# -------------------------
# Memory usage forecasting
# -------------------------
library(stats)
arima(lh, order = c(1,0,0))
arima(lh, order = c(3,0,0))
arima(lh, order = c(1,0,1))
arima(lh, order = c(3,0,0), method = "CSS")
arima(USAccDeaths, order = c(0,1,1), seasonal = list(order=c(0,1,1)))
arima(USAccDeaths, order = c(0,1,1), seasonal = list(order=c(0,1,1)),
method = "CSS") # drops first 13 observations.
# for a model with as few years as this, we want full ML
arima(LakeHuron, order = c(2,0,0), xreg = time(LakeHuron)-1920)
## presidents contains NAs
## graphs in example(acf) suggest order 1 or 3
require(graphics)
(fit1 <- arima(presidents, c(1, 0, 0)))
tsdiag(fit1)
(fit3 <- arima(presidents, c(3, 0, 0))) # smaller AIC
tsdiag(fit3)
# ----- prediction part
od <- options(digits=5) # avoid too much spurious accuracy
predict(arima(lh, order = c(3,0,0)), n.ahead = 12)
(fit <- arima(USAccDeaths, order = c(0,1,1),
seasonal = list(order=c(0,1,1))))
predict(fit, n.ahead = 6)
options(od)
# ----- Arima
library(forecast)
fit <- Arima(WWWusage,c(3,1,0))
plot(forecast(fit))
x <- fracdiff.sim( 100, ma = -.4, d = .3)$series
fit <- arfima(x)
plot(forecast(fit,h=30))
# ----- Arima forecast for memory usage (unit %) -----
library(forecast) # need to install the package "forecast"
memory.usage.threshold <- 100 # 100%
memory.usage.forecast.period <- 30 # 미래 30일분까지 예측
memory.usage.observations.startdate <- "2012-09-01"
memory.usage.observations <- c(10,11,30,35,36,39,48,56,75,69,68,72,71,72,83) # 관측치 12일분
memory.usage.period <- seq(as.Date(memory.usage.observations.startdate), length=length(memory.usage.observations), by="1 day") # 날짜세팅
memory.usage.df <- data.frame(row.names=memory.usage.period, memory=memory.usage.observations) # data.frame으로 변환
memory.usage.ts <- ts(data=memory.usage.df) # time series 생성
memory.usage.model <- auto.arima(memory.usage.ts) # arima 모델 생성
memory.usage.forecast <- forecast(memory.usage.model, h=memory.usage.forecast.period) # forecast 결과 생성
memory.usage.forecast.df <- as.data.frame(memory.usage.forecast) # forecast 결과 변환
d = memory.usage.threshold,][1,])) # 100 이 넘는 최초 데이터 추출
if(is.na(d)) {
print(sprintf("앞으로 %s일동안 %s%% 초과하지 않음", memory.usage.forecast.period, d - length(memory.usage.observations)))
} else {
print(sprintf("%s일 후에 %s%% 초과됨", d - length(memory.usage.observations), memory.usage.threshold))
}
# ---- 시각화(Plotting)
plot(memory.usage.forecast) # plotting
abline(h=100, col = "red", lty=3)
abline(v=d, col = "red", lty=3)
library(ggplot2)
library(scales)
plt <- ggplot(data=pd,aes(x=date,y=observed))
p1a<-p1a+geom_line(col='red')
p1a<-p1a+geom_line(aes(y=fitted),col='blue')
p1a<-p1a+geom_line(aes(y=forecast))+geom_ribbon(aes(ymin=lo95,ymax=hi95),alpha=.25)
p1a<-p1a+scale_x_date(name='',breaks='1 year',minor_breaks='1 month',labels=date_format("%b-%y"),expand=c(0,0))
p1a<-p1a+scale_y_continuous(name='Units of Y')
p1a<-p1a+opts(axis.text.x=theme_text(size=10),title='Arima Fit to Simulated Datan (black=forecast, blue=fitted, red=data, shadow=95% conf. interval)')
원본 소스코드는 아래에 있습니다.
https://github.com/euriion/code_snippets/blob/master/R/forecast_exam.R
