绘制图表时不显示ARIMA预测

Question

我是一个新手程序员，试图建立一个预测模型来预测苹果的股票价格。我使用这个脚本作为指导：https://github.com/inertia7/timeSeries_sp500_R/blob/master/src/script.R

我能够复制大多数部分，但当我们绘制预测数据的最后部分时，它只绘制我的训练数据集。我做错了什么？如果您有任何需要澄清的额外信息，请告诉我！

谢谢！！

​

​

# Import stock price data using existing R packages
getSymbols('AAPL', from='2018-01-01') 

# Use Augmented Dickey-Fuller Test for stationarity. To run time series models, we must adjust for non-stationarity.
adf.test(AAPL$AAPL.Adjusted)

#Augmented Dickey-Fuller Test
#data:  AAPL$AAPL.Adjusted
#Dickey-Fuller = -2.0226, Lag order = 9, p-value = 0.5687
#alternative hypothesis: stationary

#p-value is high, so we accept null hypothesis (non-stationarity). Let's visualize the data further.

# Create training set to which we'll compare values for 2021
aapl_adjusted_only = AAPL[,6] # Subset to only adjusted close price column
aapl_training = ts(aapl_adjusted_only, start=c(2018,1), end=c(2021,5), frequency = 253)

# 253 = number of trading days in a year

# Let's decompose this data set to check for any trends.
plot(decompose(aapl_training))

#From the decomposed time series of FB, we see that there is a component of seasonality here that we may need to adjust for.

#Take the difference in daily closing values to account for the non-stationary nature of the data set.
tsDiff <- diff(aapl_training)

plot(tsDiff,
     main = "Apple First Difference Time Series (2016-2021)",
     xlab = "Year",
     ylab = "Closing Values")

adf.test(tsDiff) #Run the ADF test again to test for stationary data

#Augmented Dickey-Fuller Test
#data:  tsDiff
#Dickey-Fuller = -8.6073, Lag order = 9, p-value = 0.01
#alternative hypothesis: stationary

#p-value < 0.05, so we reject the null hypothesis and accept the alternative hypothesis.


#DIAGNOSING ACF AND PACF PLOTS
ggAcf(aapl_training)
ggPacf(aapl_training)

#This means that our transformed data is stationary, and we can use the ARIMA model for forecasting.

auto.arima(aapl_training) #Select best fit ARIMA model

#Series: aapl_training 
#ARIMA(1,1,0) with drift 

#Coefficients:
#  ar1   drift
#-0.1538  0.1167
#s.e.   0.0358  0.0512

#sigma^2 estimated as 2.667:  log likelihood=-1455.84
#AIC=2917.68   AICc=2917.71   BIC=2931.59

# Now let's fit our model using the Arima method and training set
fit = Arima(aapl_training, order = c(1,1,0),
            include.drift = TRUE)

summary(fit)

#Training set error measures:
#                        ME     RMSE      MAE         MPE     MAPE       MASE        ACF1
#Training set -0.0001032566 1.629771 1.011449 -0.09118978 1.495376 0.03936189 0.003590521



# Test set that we will compare our forecast against
aapl_test = ts(aapl_adjusted_only, 
                    start = c(2021, 1), 
                    frequency = 253)

# FORECASTING
fit_arima = forecast(fit, h = 253)

if (is.null(here("models", 'arima.rds'))){
  saveRDS(fit_arima, file = here("models", 'arima.rds'))
}


forecastAAPL = autoplot(fit_arima,
                      holdout = aapl_test, 
                      forc_name = 'ARIMA', 
                      ts_object_name = 'Apple')

ggplotly(forecastAAPL)

这里是一些关于变量的str()，以便进一步澄清。

> str(aapl_adjusted_only)
 num [1:860] 41.3 41.3 41.5 42 41.8 ...
> str(aapl_test)
 Time-Series [1:860] from 2021 to 2029: 41.3 41.3 41.5 42 41.8 ...
> str(aapl_training)
 Time-Series [1:764] from 2018 to 2021: 41.3 41.3 41.5 42 41.8 ...
> str(tsDiff)
 Time-Series [1:763] from 2018 to 2021: -0.00719 0.19186 0.47243 -0.15588 -0.00479 ...
> str(fit_arima)
List of 10
 $ method   : chr "ARIMA(1,1,0) with drift"
 $ model    :List of 19
  ..$ coef     : Named num [1:2] -0.154 0.117
  .. ..- attr(*, "names")= chr [1:2] "ar1" "drift"
  ..$ sigma2   : num 2.67
  ..$ var.coef : num [1:2, 1:2] 1.28e-03 2.58e-06 2.58e-06 2.62e-03
  .. ..- attr(*, "dimnames")=List of 2
  .. .. ..$ : chr [1:2] "ar1" "drift"
  .. .. ..$ : chr [1:2] "ar1" "drift"
  ..$ mask     : logi [1:2] TRUE TRUE
  ..$ loglik   : num -1456
  ..$ aic      : num 2918
  ..$ arma     : int [1:7] 1 0 0 0 253 1 0
  ..$ residuals: Time-Series [1:764] from 2018 to 2021: 0.0412 -0.1224 0.0562 0.3673 -0.2178 ...
  ..$ call     : language Arima(y = aapl_training, order = c(1, 1, 0), include.drift = TRUE, xreg = 1:764)
  ..$ series   : chr "aapl_training"
  ..$ code     : int 0
  ..$ n.cond   : int 0
  ..$ nobs     : int 763
  ..$ model    :List of 10
  .. ..$ phi  : num -0.154
  .. ..$ theta: num(0) 
  .. ..$ Delta: num 1
  .. ..$ Z    : num [1:2] 1 1
  .. ..$ a    : num [1:2] 1.97 39.38
  .. ..$ P    : num [1:2, 1:2] 0.00 4.39e-22 4.39e-22 -4.39e-22
  .. ..$ T    : num [1:2, 1:2] -0.154 1 0 1
  .. ..$ V    : num [1:2, 1:2] 1 0 0 0
  .. ..$ h    : num 0
  .. ..$ Pn   : num [1:2, 1:2] 1.00 6.75e-23 6.75e-23 -4.39e-22
  ..$ aicc     : num 2918
  ..$ bic      : num 2932
  ..$ xreg     : int [1:764, 1] 1 2 3 4 5 6 7 8 9 10 ...
  .. ..- attr(*, "dimnames")=List of 2
  .. .. ..$ : NULL
  .. .. ..$ : chr "drift"
  ..$ x        : Time-Series [1:764] from 2018 to 2021: 41.3 41.3 41.5 42 41.8 ...
  ..$ fitted   : Time-Series [1:764] from 2018 to 2021: 41.3 41.4 41.4 41.6 42 ...
  ..- attr(*, "class")= chr [1:3] "forecast_ARIMA" "ARIMA" "Arima"
 $ level    : num [1:2] 80 95
 $ mean     : Time-Series [1:253] from 2021 to 2022: 130 130 131 131 131 ...
 $ lower    : Time-Series [1:253, 1:2] from 2021 to 2022: 128 128 127 127 127 ...
  ..- attr(*, "dimnames")=List of 2
  .. ..$ : NULL
  .. ..$ : chr [1:2] "80%" "95%"
 $ upper    : Time-Series [1:253, 1:2] from 2021 to 2022: 132 133 134 134 135 ...
  ..- attr(*, "dimnames")=List of 2
  .. ..$ : NULL
  .. ..$ : chr [1:2] "80%" "95%"
 $ x        : Time-Series [1:764] from 2018 to 2021: 41.3 41.3 41.5 42 41.8 ...
 $ series   : chr "aapl_training"
 $ fitted   : Time-Series [1:764] from 2018 to 2021: 41.3 41.4 41.4 41.6 42 ...
 $ residuals: Time-Series [1:764] from 2018 to 2021: 0.0412 -0.1224 0.0562 0.3673 -0.2178 ...
 - attr(*, "class")= chr "forecast"

Answer 1

从autoplot返回的forecastAAPL是一个ggplot对象。

您可以简单地在控制台或print(forecastAAPL)中输入forecastAAPL，它将生成预测和训练数据。