Introduction to Forecasting with ARIMA in R https://www.datascience.com/blog/introduction-to-forecasting-with-arim...

1 of 32 1/27/2019, 4:35 PM
Introduction to Forecasting with ARIMA in R https://www.datascience.com/blog/introduction-to-forecasting-with-arim...

forecast tseries ggplot2.

2 of 32 1/27/2019, 4:35 PM
Introduction to Forecasting with ARIMA in R https://www.datascience.com/blog/introduction-to-forecasting-with-arim...

tsclean()

3 of 32 1/27/2019, 4:35 PM
Introduction to Forecasting with ARIMA in R https://www.datascience.com/blog/introduction-to-forecasting-with-arim...

decompose() stl()

adf.test()

4 of 32 1/27/2019, 4:35 PM
Introduction to Forecasting with ARIMA in R https://www.datascience.com/blog/introduction-to-forecasting-with-arim...

(p, d, q)

Y
p

φ1 φ 2

d I(d)

et q

yd Y d c

5 of 32 1/27/2019, 4:35 PM
Introduction to Forecasting with ARIMA in R https://www.datascience.com/blog/introduction-to-forecasting-with-arim...

(p, d, q)

1 library('ggplot2')
2 library('forecast')
3 library('tseries')
4
5 daily_data = read.csv('day.csv', header=TRUE, stringsAsFactors=FALSE)

6 of 32 1/27/2019, 4:35 PM
Introduction to Forecasting with ARIMA in R https://www.datascience.com/blog/introduction-to-forecasting-with-arim...

1 daily_data$Date = as.Date(daily_data$dteday)
2
3 ggplot(daily_data, aes(Date, cnt)) + geom_line() + scale_x_date('month'
4 xlab("")

7 of 32 1/27/2019, 4:35 PM
Introduction to Forecasting with ARIMA in R https://www.datascience.com/blog/introduction-to-forecasting-with-arim...

1 count_ts = ts(daily_data[, c('cnt')])

2
3 daily_data$clean_cnt = tsclean(count_ts)
4
5 ggplot() +
6 geom_line(data = daily_data, aes(x = Date, y = clean_cnt)) + ylab('Cleaned Bicy

8 of 32 1/27/2019, 4:35 PM
Introduction to Forecasting with ARIMA in R https://www.datascience.com/blog/introduction-to-forecasting-with-arim...

m
Y k

m = 2k + 1

M
A(q) M A(q)

1 daily_data$cnt_ma = ma(daily_data$clean_cnt, order=7) # using the clean count

2 daily_data$cnt_ma30 = ma(daily_data$clean_cnt, order=30)
3
4
5 ggplot() +
6 geom_line(data = daily_data, aes(x = Date, y = clean_cnt, colour = "Counts"
7 geom_line(data = daily_data, aes(x = Date, y = cnt_ma, colour = "Weekly Movin
8 geom_line(data = daily_data, aes(x = Date, y = cnt_ma30, colour = "Monthly Movi
9 ylab('Bicycle Count')

9 of 32 1/27/2019, 4:35 PM
Introduction to Forecasting with ARIMA in R https://www.datascience.com/blog/introduction-to-forecasting-with-arim...

10 of 32 1/27/2019, 4:35 PM
Introduction to Forecasting with ARIMA in R https://www.datascience.com/blog/introduction-to-forecasting-with-arim...

St T E

11 of 32 1/27/2019, 4:35 PM
Introduction to Forecasting with ARIMA in R https://www.datascience.com/blog/introduction-to-forecasting-with-arim...

(P, D, Q)

stl()

1 count_ma = ts(na.omit(daily_data$cnt_ma), frequency=30)

2 decomp = stl(count_ma, s.window="periodic")
3 deseasonal_cnt <- seasadj(decomp)
4 plot(decomp)

stl()
allow.multiplicative.trend=TRUE

seasadj()
forecast

ts()

12 of 32 1/27/2019, 4:35 PM
Introduction to Forecasting with ARIMA in R https://www.datascience.com/blog/introduction-to-forecasting-with-arim...

13 of 32 1/27/2019, 4:35 PM
Introduction to Forecasting with ARIMA in R https://www.datascience.com/blog/introduction-to-forecasting-with-arim...

14 of 32 1/27/2019, 4:35 PM
Introduction to Forecasting with ARIMA in R https://www.datascience.com/blog/introduction-to-forecasting-with-arim...

1 adf.test(count_ma, alternative = "stationary")

2
3 Augmented Dickey-Fuller Test
4
5 data: count_ma
6 Dickey-Fuller = -0.2557, Lag order = 8, p-value = 0.99
7 alternative hypothesis: stationary

(d = 2)

15 of 32 1/27/2019, 4:35 PM
Introduction to Forecasting with ARIMA in R https://www.datascience.com/blog/introduction-to-forecasting-with-arim...

M A (q)

AR(p)

16 of 32 1/27/2019, 4:35 PM
Introduction to Forecasting with ARIMA in R https://www.datascience.com/blog/introduction-to-forecasting-with-arim...

1 Acf(count_ma, main='')
2
3 Pacf(count_ma, main='')

d=1

17 of 32 1/27/2019, 4:35 PM
Introduction to Forecasting with ARIMA in R https://www.datascience.com/blog/introduction-to-forecasting-with-arim...

1 count_d1 = diff(deseasonal_cnt, differences = 1)

2 plot(count_d1)
3 adf.test(count_d1, alternative = "stationary")
4
5 Augmented Dickey-Fuller Test
6
7 data: count_d1
8 Dickey-Fuller = -9.9255, Lag order = 8, p-value = 0.01
9 alternative hypothesis: stationary

p q

1 Acf(count_d1, main='ACF for Differenced Series')

2 Pacf(count_d1, main='PACF for Differenced Series')

18 of 32 1/27/2019, 4:35 PM