Прогнозирование с использованием модели GARCH в R - PullRequest
Новая
       7

Прогнозирование с использованием модели GARCH в R

0 голосов
/ 20 января 2019 
df=structure(list(X = c(1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L), 
    json_data.time.updated = structure(1:41, .Label = c("Jan 19, 2019 15:18:00 UTC", 
    "Jan 19, 2019 15:19:00 UTC", "Jan 19, 2019 15:51:00 UTC", 
    "Jan 19, 2019 15:52:00 UTC", "Jan 19, 2019 15:54:00 UTC", 
    "Jan 19, 2019 15:55:00 UTC", "Jan 19, 2019 15:57:00 UTC", 
    "Jan 19, 2019 15:58:00 UTC", "Jan 19, 2019 16:00:00 UTC", 
    "Jan 19, 2019 16:01:00 UTC", "Jan 19, 2019 16:03:00 UTC", 
    "Jan 19, 2019 16:04:00 UTC", "Jan 19, 2019 16:06:00 UTC", 
    "Jan 19, 2019 16:07:00 UTC", "Jan 19, 2019 16:09:00 UTC", 
    "Jan 19, 2019 16:10:00 UTC", "Jan 19, 2019 16:12:00 UTC", 
    "Jan 19, 2019 16:13:00 UTC", "Jan 19, 2019 16:15:00 UTC", 
    "Jan 19, 2019 16:16:00 UTC", "Jan 19, 2019 16:18:00 UTC", 
    "Jan 19, 2019 16:19:00 UTC", "Jan 19, 2019 16:21:00 UTC", 
    "Jan 19, 2019 16:22:00 UTC", "Jan 19, 2019 16:24:00 UTC", 
    "Jan 19, 2019 16:25:00 UTC", "Jan 19, 2019 16:27:00 UTC", 
    "Jan 19, 2019 16:28:00 UTC", "Jan 19, 2019 16:30:00 UTC", 
    "Jan 19, 2019 16:31:00 UTC", "Jan 19, 2019 16:33:00 UTC", 
    "Jan 19, 2019 16:34:00 UTC", "Jan 19, 2019 16:36:00 UTC", 
    "Jan 19, 2019 16:37:00 UTC", "Jan 19, 2019 16:39:00 UTC", 
    "Jan 19, 2019 16:40:00 UTC", "Jan 19, 2019 16:42:00 UTC", 
    "Jan 19, 2019 16:43:00 UTC", "Jan 19, 2019 16:45:00 UTC", 
    "Jan 19, 2019 16:46:00 UTC", "Jan 19, 2019 16:48:00 UTC"), class = "factor"), 
    json_data.time.updatedISO = structure(1:41, .Label = c("2019-01-19T15:18:00+00:00", 
    "2019-01-19T15:19:00+00:00", "2019-01-19T15:51:00+00:00", 
    "2019-01-19T15:52:00+00:00", "2019-01-19T15:54:00+00:00", 
    "2019-01-19T15:55:00+00:00", "2019-01-19T15:57:00+00:00", 
    "2019-01-19T15:58:00+00:00", "2019-01-19T16:00:00+00:00", 
    "2019-01-19T16:01:00+00:00", "2019-01-19T16:03:00+00:00", 
    "2019-01-19T16:04:00+00:00", "2019-01-19T16:06:00+00:00", 
    "2019-01-19T16:07:00+00:00", "2019-01-19T16:09:00+00:00", 
    "2019-01-19T16:10:00+00:00", "2019-01-19T16:12:00+00:00", 
    "2019-01-19T16:13:00+00:00", "2019-01-19T16:15:00+00:00", 
    "2019-01-19T16:16:00+00:00", "2019-01-19T16:18:00+00:00", 
    "2019-01-19T16:19:00+00:00", "2019-01-19T16:21:00+00:00", 
    "2019-01-19T16:22:00+00:00", "2019-01-19T16:24:00+00:00", 
    "2019-01-19T16:25:00+00:00", "2019-01-19T16:27:00+00:00", 
    "2019-01-19T16:28:00+00:00", "2019-01-19T16:30:00+00:00", 
    "2019-01-19T16:31:00+00:00", "2019-01-19T16:33:00+00:00", 
    "2019-01-19T16:34:00+00:00", "2019-01-19T16:36:00+00:00", 
    "2019-01-19T16:37:00+00:00", "2019-01-19T16:39:00+00:00", 
    "2019-01-19T16:40:00+00:00", "2019-01-19T16:42:00+00:00", 
    "2019-01-19T16:43:00+00:00", "2019-01-19T16:45:00+00:00", 
    "2019-01-19T16:46:00+00:00", "2019-01-19T16:48:00+00:00"), class = "factor"), 
    json_data.time.updateduk = structure(1:41, .Label = c("Jan 19, 2019 at 15:18 GMT", 
    "Jan 19, 2019 at 15:19 GMT", "Jan 19, 2019 at 15:51 GMT", 
    "Jan 19, 2019 at 15:52 GMT", "Jan 19, 2019 at 15:54 GMT", 
    "Jan 19, 2019 at 15:55 GMT", "Jan 19, 2019 at 15:57 GMT", 
    "Jan 19, 2019 at 15:58 GMT", "Jan 19, 2019 at 16:00 GMT", 
    "Jan 19, 2019 at 16:01 GMT", "Jan 19, 2019 at 16:03 GMT", 
    "Jan 19, 2019 at 16:04 GMT", "Jan 19, 2019 at 16:06 GMT", 
    "Jan 19, 2019 at 16:07 GMT", "Jan 19, 2019 at 16:09 GMT", 
    "Jan 19, 2019 at 16:10 GMT", "Jan 19, 2019 at 16:12 GMT", 
    "Jan 19, 2019 at 16:13 GMT", "Jan 19, 2019 at 16:15 GMT", 
    "Jan 19, 2019 at 16:16 GMT", "Jan 19, 2019 at 16:18 GMT", 
    "Jan 19, 2019 at 16:19 GMT", "Jan 19, 2019 at 16:21 GMT", 
    "Jan 19, 2019 at 16:22 GMT", "Jan 19, 2019 at 16:24 GMT", 
    "Jan 19, 2019 at 16:25 GMT", "Jan 19, 2019 at 16:27 GMT", 
    "Jan 19, 2019 at 16:28 GMT", "Jan 19, 2019 at 16:30 GMT", 
    "Jan 19, 2019 at 16:31 GMT", "Jan 19, 2019 at 16:33 GMT", 
    "Jan 19, 2019 at 16:34 GMT", "Jan 19, 2019 at 16:36 GMT", 
    "Jan 19, 2019 at 16:37 GMT", "Jan 19, 2019 at 16:39 GMT", 
    "Jan 19, 2019 at 16:40 GMT", "Jan 19, 2019 at 16:42 GMT", 
    "Jan 19, 2019 at 16:43 GMT", "Jan 19, 2019 at 16:45 GMT", 
    "Jan 19, 2019 at 16:46 GMT", "Jan 19, 2019 at 16:48 GMT"), class = "factor"), 
    code = structure(c(1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
    1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
    1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
    1L), .Label = "USD", class = "factor"), rate = structure(1:41, .Label = c("3,735.7750", 
    "3,735.9150", "3,736.9100", "3,735.3200", "3,736.7717", "3,736.0750", 
    "3,734.9600", "3,734.9117", "3,734.2833", "3,734.4950", "3,735.8533", 
    "3,736.1917", "3,735.5450", "3,735.5867", "3,736.0617", "3,736.3417", 
    "3,737.0633", "3,736.9583", "3,737.1667", "3,737.1433", "3,737.0583", 
    "3,736.9283", "3,737.6383", "3,737.5167", "3,737.9133", "3,738.7533", 
    "3,738.6767", "3,738.5767", "3,738.5917", "3,738.8867", "3,739.6333", 
    "3,739.9600", "3,739.3383", "3,739.9267", "3,739.3067", "3,739.5867", 
    "3,739.6567", "3,739.4267", "3,739.1500", "3,739.8817", "3,739.5550"
    ), class = "factor"), description = structure(c(1L, 1L, 1L, 
    1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
    1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
    1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L), .Label = "United States Dollar", class = "factor"), 
    rate_float = structure(1:41, .Label = c("3735.775", "3735.915", 
    "3736.91", "3735.32", "3736.7717", "3736.075", "3734.96", 
    "3734.9117", "3734.2833", "3734.495", "3735.8533", "3736.1917", 
    "3735.545", "3735.5867", "3736.0617", "3736.3417", "3737.0633", 
    "3736.9583", "3737.1667", "3737.1433", "3737.0583", "3736.9283", 
    "3737.6383", "3737.5167", "3737.9133", "3738.7533", "3738.6767", 
    "3738.5767", "3738.5917", "3738.8867", "3739.6333", "3739.96", 
    "3739.3383", "3739.9267", "3739.3067", "3739.5867", "3739.6567", 
    "3739.4267", "3739.15", "3739.8817", "3739.555"), class = "factor")), row.names = c(NA, 
41L), class = "data.frame")

Я хочу выполнить модель GARCH, поэтому у меня есть этот код 

df$rate_float=as.numeric(df$rate_float)
library("fGarch")
b=garchFit(formula = ~ garch(1, 1), data = df$rate_float, 
         init.rec = c( "mci","uev"), 
         delta = 2, skew = 1, shape = 4, 
         cond.dist = c("norm", "snorm", "ged", "sged", "std", "sstd", 
                       "snig", "QMLE"), 
         include.mean = TRUE, include.delta = NULL, include.skew = NULL, 
         include.shape = NULL, leverage = NULL, trace = TRUE, 

         algorithm = c("nlminb", "lbfgsb", "nlminb+nm", "lbfgsb+nm"), 
         hessian = c("ropt"), control = list(), 
         title = NULL, description = NULL)

garchKappa(cond.dist = c("norm", "ged", "std", "snorm", "sged", "sstd",
                         "snig"), gamma = 0, delta = 2, skew = NA, shape = NA)

Таким образом, переменная времени - json_data.time.updateduk. Как я могу создать прогноз на 10 минут вперед? Второй, но связанный вопрос, как я могу получить прогноз для начальных значений, чтобы понять качество прогноза так как выходной мне нужно 1. прогноз на 10 минут вперед, как это enter image description here

  1. Остаточные И.Е. что-то вроде этого enter image description here

Как это сделать?

1 Ответ

0 голосов
/ 20 января 2019

Вы можете прогнозировать на 10 шагов вперед с помощью 

predict(b, n.ahead = 10)
#    meanForecast meanError standardDeviation
# 1      20.19952  20.82351          20.82351
# 2      20.19952  20.84651          20.84651
# 3      20.19952  20.86948          20.86948
# 4      20.19952  20.89243          20.89243
# 5      20.19952  20.91535          20.91535
# 6      20.19952  20.93825          20.93825
# 7      20.19952  20.96112          20.96112
# 8      20.19952  20.98396          20.98396
# 9      20.19952  21.00679          21.00679
# 10     20.19952  21.02959          21.02959

, в то время как подогнанные значения задаются как 

fitted(b)
#        1        2        3        4        5        6        7        8        9       10 
# 20.19952 20.19952 20.19952 20.19952 20.19952 20.19952 20.19952 20.19952 20.19952 20.19952 
#       11       12       13       14       15       16       17       18       19       20 
# 20.19952 20.19952 20.19952 20.19952 20.19952 20.19952 20.19952 20.19952 20.19952 20.19952 
#       21       22       23       24       25       26       27       28       29       30 
# 20.19952 20.19952 20.19952 20.19952 20.19952 20.19952 20.19952 20.19952 20.19952 20.19952 
#       31       32       33       34       35       36       37       38       39       40 
# 20.19952 20.19952 20.19952 20.19952 20.19952 20.19952 20.19952 20.19952 20.19952 20.19952 
#       41 
# 20.19952

Средние прогнозы и подогнанные значения постоянны, так как вы моделируете условную дисперсию;по этой причине две другие колонки прогнозов непостоянны.Установленные условные стандартные отклонения задаются как 

b@sigma.t
#  [1] 11.8995423 19.2244571 18.2258257 17.2273530 16.2290683 15.2310087 14.2332215 13.2357682
#  [9] 12.2387307 11.2422194 10.2463878  9.2514556  8.2577478  7.2657667  6.2763312  5.2908696
# [17]  4.3121073  3.3459283  2.4075307  1.5482802  0.9990692  1.2645580  2.0494083  2.9666550
# [25]  3.9245394  4.8992833  5.8825118  6.8705831  7.8616716  8.8547641  9.8492545 10.8447583
# [33] 11.8410198 12.8378627 13.8351612 14.8328235 15.8307808 16.8289806 17.8273822 18.8259534
# [41] 19.8246687
Добро пожаловать на сайт PullRequest, где вы можете задавать вопросы и получать ответы от других членов сообщества.

Нет похожих вопросов

...