title: “Tugas ARMA”
author: “Grace Maria Alua (240611010050)”
date: “2026-11-19”
output: html_document
Kode berikut memuat paket yang diperlukan untuk analisis deret waktu
(time series), membaca file
P_Data_Extract_From_World_Development_Indicators.xlsx
langsung dari folder Downloads, serta memfilter seluruh
indikator ekonomi dan demografi milik Jepang
(Country Code == "JPN").
library(readxl)
library(dplyr)
library(tidyr)
library(tseries)
library(forecast)
library(vars)
library(knitr)
# Membaca file Excel langsung dari folder Downloads pengguna Windows
file_path <- file.path(Sys.getenv("USERPROFILE"), "Downloads", "P_Data_Extract_From_World_Development_Indicators.xlsx")
df_raw <- read_excel(file_path, sheet = "Data")
# Filter khusus negara Jepang (JPN)
df_jpn <- df_raw %>%
filter(`Country Code` == "JPN")
# Mengubah format wide (tahun sebagai kolom) menjadi format long lalu wide berdasarkan tahun
df_long <- df_jpn %>%
dplyr::select(Indikator = A, matches("\\[YR")) %>%
pivot_longer(
cols = matches("\\[YR"),
names_to = "Tahun_Str",
values_to = "Nilai_Str"
) %>%
mutate(
Tahun = as.integer(substr(Tahun_Str, 1, 4)),
Nilai = suppressWarnings(as.numeric(Nilai_Str))
) %>%
dplyr::select(Tahun, Indikator, Nilai) %>%
arrange(Indikator, Tahun)
df_wide <- df_long %>%
pivot_wider(names_from = Indikator, values_from = Nilai) %>%
arrange(Tahun)
kable(df_wide, digits = 4, caption = "Data Deret Waktu Jepang (2001 - 2025)")
| Tahun | Exports of goods and services (% of GDP) | Foreign direct investment, net inflows (% of GDP) | Imports of goods and services (% of GDP) | Population growth (annual %) |
|---|---|---|---|---|
| 2001 | 9.9313 | 0.1110 | 9.3459 | 0.2410 |
| 2002 | 10.6941 | 0.2722 | 9.4491 | 0.2325 |
| 2003 | 11.3308 | 0.1918 | 9.7442 | 0.2140 |
| 2004 | 12.6608 | 0.1523 | 10.7716 | 0.0337 |
| 2005 | 13.6894 | 0.1120 | 12.3016 | 0.0094 |
| 2006 | 15.4992 | -0.0516 | 14.2157 | 0.0634 |
| 2007 | 17.1191 | 0.4677 | 15.3766 | 0.1149 |
| 2008 | 17.0551 | 0.4772 | 16.7187 | 0.0484 |
| 2009 | 12.3085 | 0.2291 | 11.8646 | -0.0125 |
| 2010 | 14.7827 | 0.1280 | 13.4573 | 0.0180 |
| 2011 | 14.6645 | -0.0135 | 15.3069 | -0.1852 |
| 2012 | 14.2739 | 0.0086 | 15.9004 | -0.1597 |
| 2013 | 15.6037 | 0.2020 | 17.9877 | -0.1443 |
| 2014 | 17.1079 | 0.3962 | 19.6565 | -0.1327 |
| 2015 | 17.0921 | 0.1158 | 17.6355 | -0.1061 |
| 2016 | 15.7224 | 0.8014 | 14.9339 | -0.0511 |
| 2017 | 17.2158 | 0.3732 | 16.4730 | -0.0819 |
| 2018 | 17.9111 | 0.4906 | 17.8924 | -0.1269 |
| 2019 | 17.0378 | 0.7618 | 17.3205 | -0.1405 |
| 2020 | 15.1287 | 1.2061 | 15.3905 | -0.2942 |
| 2021 | 17.4620 | 0.6703 | 17.9634 | -0.4600 |
| 2022 | 20.6479 | 1.0826 | 24.2128 | -0.4439 |
| 2023 | 21.0593 | 0.4563 | 22.5358 | -0.4874 |
| 2024 | 21.9799 | 0.4505 | 22.8730 | -0.4357 |
| 2025 | NA | 1.0328 | NA | -0.4921 |
Model AR, MA, dan ARMA mensyaratkan data dalam kondisi stasioner. Jika data tidak stasioner pada tingkat level (\(p\text{-value} > 0,05\)), maka diperlukan proses differencing (\(d = 1\)) sehingga model berkembang menjadi ARIMA(\(p,d,q\)).
indikator_list <- unique(df_long$Indikator)
hasil_adf <- lapply(indikator_list, function(ind) {
s <- df_long %>%
filter(Indikator == ind, !is.na(Nilai)) %>%
pull(Nilai)
adf_lvl <- adf.test(s)
adf_diff1 <- adf.test(diff(s))
data.frame(
Indikator = ind,
N = length(s),
Mean = round(mean(s), 4),
SD = round(sd(s), 4),
ADF_Level = round(adf_lvl$statistic, 4),
P_Val_Level = round(adf_lvl$p.value, 4),
ADF_Diff1 = round(adf_diff1$statistic, 4),
P_Val_Diff1 = round(adf_diff1$p.value, 4),
Kesimpulan = ifelse(adf_lvl$p.value <= 0.05, "Stasioner di Level", "Stasioner di 1st Difference (I(1))")
)
}) %>% bind_rows()
kable(hasil_adf, row.names = FALSE, caption = "Statistik Deskriptif dan Uji Augmented Dickey-Fuller (ADF)")
| Indikator | N | Mean | SD | ADF_Level | P_Val_Level | ADF_Diff1 | P_Val_Diff1 | Kesimpulan |
|---|---|---|---|---|---|---|---|---|
| Exports of goods and services (% of GDP) | 24 | 15.7491 | 3.0855 | -2.5113 | 0.3776 | -2.7840 | 0.2737 | Stasioner di 1st Difference (I(1)) |
| Foreign direct investment, net inflows (% of GDP) | 25 | 0.4050 | 0.3503 | -2.4073 | 0.4172 | -3.1978 | 0.1161 | Stasioner di 1st Difference (I(1)) |
| Imports of goods and services (% of GDP) | 24 | 15.8053 | 4.0801 | -2.4816 | 0.3889 | -2.9456 | 0.2122 | Stasioner di 1st Difference (I(1)) |
| Population growth (annual %) | 25 | -0.1112 | 0.2226 | -2.4659 | 0.3949 | -2.9388 | 0.2147 | Stasioner di 1st Difference (I(1)) |
par(mfrow = c(4, 3), mar = c(3, 3, 2.5, 1))
for (ind in indikator_list) {
sub_df <- df_long %>% filter(Indikator == ind, !is.na(Nilai))
ts_obj <- ts(sub_df$Nilai, start = min(sub_df$Tahun), frequency = 1)
plot(ts_obj, main = paste("Tren:", substr(ind, 1, 25)), ylab = "%", xlab = "Tahun", col = "navy", lwd = 2)
acf(diff(ts_obj), main = paste("ACF Diff1:", substr(ind, 1, 18)))
pacf(diff(ts_obj), main = paste("PACF Diff1:", substr(ind, 1, 18)))
}
par(mfrow = c(1, 1))
Untuk mengevaluasi performa prediksi tiap model secara objektif, data
dibagi menjadi data latih (train) dan
4 tahun terakhir sebagai data uji (test).
Estimasi menggunakan metode Maximum Likelihood
(method = "ML") agar proses komputasi berjalan cepat dan
stabil.
evaluasi_semua <- list()
model_terbaik_list <- list()
for (ind in indikator_list) {
sub_df <- df_long %>% filter(Indikator == ind, !is.na(Nilai))
n_total <- nrow(sub_df)
n_test <- 4
n_train <- n_total - n_test
ts_train <- ts(sub_df$Nilai[1:n_train], start = sub_df$Tahun[1], frequency = 1)
actual_test <- sub_df$Nilai[(n_train + 1):n_total]
kandidat <- list(
"AR(1)" = tryCatch(Arima(ts_train, order = c(1, 0, 0), method = "ML"), error = function(e) NULL),
"MA(1)" = tryCatch(Arima(ts_train, order = c(0, 0, 1), method = "ML"), error = function(e) NULL),
"ARMA(1,1)" = tryCatch(Arima(ts_train, order = c(1, 0, 1), method = "ML"), error = function(e) NULL),
"ARIMA(1,1,0)" = tryCatch(Arima(ts_train, order = c(1, 1, 0), method = "ML"), error = function(e) NULL),
"ARIMA(0,1,1)" = tryCatch(Arima(ts_train, order = c(0, 1, 1), method = "ML"), error = function(e) NULL),
"ARIMA(1,1,1)" = tryCatch(Arima(ts_train, order = c(1, 1, 1), method = "ML"), error = function(e) NULL),
"Holt Linear Trend" = tryCatch(holt(ts_train, h = n_test), error = function(e) NULL)
)
res_ind <- lapply(names(kandidat), function(nama_mod) {
mod <- kandidat[[nama_mod]]
if (is.null(mod)) return(NULL)
fc <- forecast(mod, h = n_test)
pred <- as.numeric(fc$mean)
rmse_val <- sqrt(mean((actual_test - pred)^2))
mae_val <- mean(abs(actual_test - pred))
aic_val <- if (!is.null(mod$aic)) round(mod$aic, 2) else round(mod$model$aic, 2)
bic_val <- if (!is.null(mod$bic)) round(mod$bic, 2) else round(mod$model$bic, 2)
data.frame(
Model = nama_mod,
AIC = aic_val,
BIC = bic_val,
RMSE = round(rmse_val, 4),
MAE = round(mae_val, 4)
)
}) %>% bind_rows() %>% arrange(RMSE)
evaluasi_semua[[ind]] <- res_ind
model_terbaik_list[[ind]] <- res_ind$Model[1]
}
for (ind in indikator_list) {
cat("\n### Evaluasi Model:", ind, "\n")
print(kable(evaluasi_semua[[ind]], row.names = FALSE))
}
##
## ### Evaluasi Model: Exports of goods and services (% of GDP)
##
##
## |Model | AIC| BIC| RMSE| MAE|
## |:-----------------|-----:|-----:|------:|------:|
## |Holt Linear Trend | 88.17| 93.15| 4.3312| 4.1183|
## |ARIMA(1,1,1) | 78.51| 81.35| 5.0820| 4.8192|
## |ARIMA(0,1,1) | 76.80| 78.69| 5.2699| 4.9879|
## |ARIMA(1,1,0) | 76.83| 78.72| 5.3364| 5.0573|
## |MA(1) | 86.49| 89.47| 5.8464| 5.6221|
## |AR(1) | 81.68| 84.66| 5.8465| 5.5522|
## |ARMA(1,1) | 83.52| 87.51| 5.9670| 5.6895|
##
## ### Evaluasi Model: Foreign direct investment, net inflows (% of GDP)
##
##
## |Model | AIC| BIC| RMSE| MAE|
## |:-----------------|-----:|-----:|------:|------:|
## |ARIMA(1,1,0) | 9.05| 11.04| 0.2959| 0.2819|
## |ARIMA(0,1,1) | 7.95| 9.94| 0.3043| 0.3022|
## |ARIMA(1,1,1) | 9.95| 12.94| 0.3043| 0.3021|
## |Holt Linear Trend | 15.01| 20.23| 0.3291| 0.3022|
## |ARMA(1,1) | 11.03| 15.21| 0.3462| 0.2968|
## |AR(1) | 9.59| 12.72| 0.4383| 0.3289|
## |MA(1) | 11.52| 14.65| 0.5119| 0.4162|
##
## ### Evaluasi Model: Imports of goods and services (% of GDP)
##
##
## |Model | AIC| BIC| RMSE| MAE|
## |:-----------------|-----:|-----:|------:|------:|
## |Holt Linear Trend | 93.56| 98.54| 6.0965| 5.7040|
## |ARIMA(1,1,0) | 81.85| 83.74| 7.0198| 6.6120|
## |ARIMA(1,1,1) | 83.39| 86.23| 7.0581| 6.6671|
## |ARIMA(0,1,1) | 81.79| 83.67| 7.1190| 6.7179|
## |AR(1) | 87.34| 90.33| 7.4959| 7.0672|
## |MA(1) | 93.53| 96.52| 7.8049| 7.4497|
## |ARMA(1,1) | 88.43| 92.41| 7.8268| 7.4279|
##
## ### Evaluasi Model: Population growth (annual %)
##
##
## |Model | AIC| BIC| RMSE| MAE|
## |:-----------------|------:|------:|------:|------:|
## |ARIMA(1,1,1) | -36.10| -33.11| 0.0308| 0.0258|
## |ARIMA(0,1,1) | -38.10| -36.11| 0.0344| 0.0254|
## |ARIMA(1,1,0) | -38.09| -36.09| 0.0435| 0.0363|
## |AR(1) | -35.26| -32.12| 0.0634| 0.0509|
## |ARMA(1,1) | -34.27| -30.09| 0.0678| 0.0567|
## |Holt Linear Trend | -31.62| -26.39| 0.1209| 0.1175|
## |MA(1) | -23.79| -20.65| 0.3795| 0.3638|
Model dengan nilai RMSE terendah untuk masing-masing indikator dilatih kembali menggunakan seluruh data historis untuk memproyeksikan 5 periode ke depan.
par(mfrow = c(2, 2), mar = c(4, 4, 3, 1))
for (ind in indikator_list) {
sub_df <- df_long %>% filter(Indikator == ind, !is.na(Nilai))
ts_full <- ts(sub_df$Nilai, start = min(sub_df$Tahun), frequency = 1)
terbaik <- model_terbaik_list[[ind]]
fit_final <- switch(
terbaik,
"AR(1)" = Arima(ts_full, order = c(1, 0, 0), method = "ML"),
"MA(1)" = Arima(ts_full, order = c(0, 0, 1), method = "ML"),
"ARMA(1,1)" = Arima(ts_full, order = c(1, 0, 1), method = "ML"),
"ARIMA(1,1,0)" = Arima(ts_full, order = c(1, 1, 0), method = "ML"),
"ARIMA(0,1,1)" = Arima(ts_full, order = c(0, 1, 1), method = "ML"),
"ARIMA(1,1,1)" = Arima(ts_full, order = c(1, 1, 1), method = "ML"),
"Holt Linear Trend" = holt(ts_full, h = 5)
)
fc_final <- forecast(fit_final, h = 5)
plot(
fc_final,
main = paste0(substr(ind, 1, 28), "\n(", terbaik, ")"),
xlab = "Tahun",
ylab = "Nilai (%)",
col = "navy",
lwd = 2
)
grid()
cat("\n**Proyeksi 5 Periode ke Depan -", ind, "(Model:", terbaik, ")**\n")
print(kable(as.data.frame(fc_final), digits = 4))
}
##
## **Proyeksi 5 Periode ke Depan - Exports of goods and services (% of GDP) (Model: Holt Linear Trend )**
##
##
## | | Point Forecast| Lo 80| Hi 80| Lo 95| Hi 95|
## |:----|--------------:|-------:|-------:|-------:|-------:|
## |2025 | 22.4256| 20.1826| 24.6686| 18.9953| 25.8559|
## |2026 | 22.9443| 20.0138| 25.8747| 18.4625| 27.4260|
## |2027 | 23.4629| 19.9780| 26.9479| 18.1331| 28.7927|
## |2028 | 23.9816| 20.0189| 27.9444| 17.9211| 30.0421|
## |2029 | 24.5003| 20.1114| 28.8892| 17.7881| 31.2125|
##
## **Proyeksi 5 Periode ke Depan - Foreign direct investment, net inflows (% of GDP) (Model: ARIMA(1,1,0) )**
##
##
## | | Point Forecast| Lo 80| Hi 80| Lo 95| Hi 95|
## |:----|--------------:|------:|------:|-------:|------:|
## |2026 | 0.7573| 0.3673| 1.1472| 0.1608| 1.3537|
## |2027 | 0.8876| 0.4468| 1.3284| 0.2135| 1.5618|
## |2028 | 0.8259| 0.2968| 1.3551| 0.0167| 1.6352|
## |2029 | 0.8551| 0.2693| 1.4410| -0.0409| 1.7511|
## |2030 | 0.8413| 0.1958| 1.4868| -0.1459| 1.8285|
##
## **Proyeksi 5 Periode ke Depan - Imports of goods and services (% of GDP) (Model: Holt Linear Trend )**
##
##
## | | Point Forecast| Lo 80| Hi 80| Lo 95| Hi 95|
## |:----|--------------:|-------:|-------:|-------:|-------:|
## |2025 | 21.9280| 19.1421| 24.7138| 17.6674| 26.1886|
## |2026 | 22.4121| 19.6262| 25.1979| 18.1515| 26.6726|
## |2027 | 22.8962| 20.1103| 25.6820| 18.6356| 27.1567|
## |2028 | 23.3803| 20.5944| 26.1661| 19.1197| 27.6408|
## |2029 | 23.8644| 21.0785| 26.6502| 19.6038| 28.1249|
##
## **Proyeksi 5 Periode ke Depan - Population growth (annual %) (Model: ARIMA(1,1,1) )**
##
##
## | | Point Forecast| Lo 80| Hi 80| Lo 95| Hi 95|
## |:----|--------------:|-------:|-------:|-------:|-------:|
## |2026 | -0.4972| -0.6045| -0.3900| -0.6612| -0.3333|
## |2027 | -0.4994| -0.6596| -0.3392| -0.7444| -0.2544|
## |2028 | -0.5003| -0.7030| -0.2977| -0.8103| -0.1904|
## |2029 | -0.5007| -0.7395| -0.2620| -0.8658| -0.1356|
## |2030 | -0.5009| -0.7714| -0.2304| -0.9145| -0.0872|
par(mfrow = c(1, 1))
Untuk menganalisis keterkaitan dinamis antar keempat indikator ekonomi dan demografi Jepang secara simultan pada periode 2001–2024, digunakan model VAR(1) pada data first difference yang telah stasioner.
df_var_raw <- df_wide %>%
filter(Tahun <= 2024) %>%
dplyr::select(-Tahun)
# Differencing pertama agar seluruh variabel stasioner
df_var_diff <- as.data.frame(lapply(df_var_raw, diff))
colnames(df_var_diff) <- c("FDI_GDP", "Exports_GDP", "Imports_GDP", "Pop_Growth")
# Estimasi model VAR(1)
model_var <- VAR(df_var_diff, p = 1, type = "const")
summary(model_var)
##
## VAR Estimation Results:
## =========================
## Endogenous variables: FDI_GDP, Exports_GDP, Imports_GDP, Pop_Growth
## Deterministic variables: const
## Sample size: 22
## Log Likelihood: -44.622
## Roots of the characteristic polynomial:
## 0.423 0.3812 0.3812 0.1688
## Call:
## VAR(y = df_var_diff, p = 1, type = "const")
##
##
## Estimation results for equation FDI_GDP:
## ========================================
## FDI_GDP = FDI_GDP.l1 + Exports_GDP.l1 + Imports_GDP.l1 + Pop_Growth.l1 + const
##
## Estimate Std. Error t value Pr(>|t|)
## FDI_GDP.l1 0.2626 0.5294 0.496 0.626
## Exports_GDP.l1 -0.1609 1.3882 -0.116 0.909
## Imports_GDP.l1 -0.2648 0.3913 -0.677 0.508
## Pop_Growth.l1 -3.3212 6.1002 -0.544 0.593
## const 0.4315 0.4774 0.904 0.379
##
##
## Residual standard error: 1.858 on 17 degrees of freedom
## Multiple R-Squared: 0.05045, Adjusted R-squared: -0.173
## F-statistic: 0.2258 on 4 and 17 DF, p-value: 0.9202
##
##
## Estimation results for equation Exports_GDP:
## ============================================
## Exports_GDP = FDI_GDP.l1 + Exports_GDP.l1 + Imports_GDP.l1 + Pop_Growth.l1 + const
##
## Estimate Std. Error t value Pr(>|t|)
## FDI_GDP.l1 0.02642 0.08777 0.301 0.7670
## Exports_GDP.l1 -0.54376 0.23014 -2.363 0.0303 *
## Imports_GDP.l1 -0.02315 0.06487 -0.357 0.7256
## Pop_Growth.l1 0.98653 1.01131 0.975 0.3430
## const 0.04981 0.07915 0.629 0.5375
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
##
## Residual standard error: 0.308 on 17 degrees of freedom
## Multiple R-Squared: 0.3102, Adjusted R-squared: 0.1479
## F-statistic: 1.911 on 4 and 17 DF, p-value: 0.1549
##
##
## Estimation results for equation Imports_GDP:
## ============================================
## Imports_GDP = FDI_GDP.l1 + Exports_GDP.l1 + Imports_GDP.l1 + Pop_Growth.l1 + const
##
## Estimate Std. Error t value Pr(>|t|)
## FDI_GDP.l1 0.68950 0.66368 1.039 0.313
## Exports_GDP.l1 -0.09583 1.74015 -0.055 0.957
## Imports_GDP.l1 -0.48342 0.49051 -0.986 0.338
## Pop_Growth.l1 -10.23538 7.64676 -1.339 0.198
## const 0.21390 0.59848 0.357 0.725
##
##
## Residual standard error: 2.329 on 17 degrees of freedom
## Multiple R-Squared: 0.1425, Adjusted R-squared: -0.05928
## F-statistic: 0.7062 on 4 and 17 DF, p-value: 0.5986
##
##
## Estimation results for equation Pop_Growth:
## ===========================================
## Pop_Growth = FDI_GDP.l1 + Exports_GDP.l1 + Imports_GDP.l1 + Pop_Growth.l1 + const
##
## Estimate Std. Error t value Pr(>|t|)
## FDI_GDP.l1 -0.01826 0.02296 -0.795 0.4375
## Exports_GDP.l1 -0.11203 0.06021 -1.861 0.0802 .
## Imports_GDP.l1 0.01249 0.01697 0.736 0.4717
## Pop_Growth.l1 0.21130 0.26456 0.799 0.4355
## const -0.01987 0.02071 -0.960 0.3507
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
##
## Residual standard error: 0.08058 on 17 degrees of freedom
## Multiple R-Squared: 0.1695, Adjusted R-squared: -0.02593
## F-statistic: 0.8673 on 4 and 17 DF, p-value: 0.5034
##
##
##
## Covariance matrix of residuals:
## FDI_GDP Exports_GDP Imports_GDP Pop_Growth
## FDI_GDP 3.45213 -0.093343 3.767686 0.028459
## Exports_GDP -0.09334 0.094878 0.008438 0.003566
## Imports_GDP 3.76769 0.008438 5.424435 0.001952
## Pop_Growth 0.02846 0.003566 0.001952 0.006493
##
## Correlation matrix of residuals:
## FDI_GDP Exports_GDP Imports_GDP Pop_Growth
## FDI_GDP 1.0000 -0.16310 0.87067 0.1901
## Exports_GDP -0.1631 1.00000 0.01176 0.1437
## Imports_GDP 0.8707 0.01176 1.00000 0.0104
## Pop_Growth 0.1901 0.14366 0.01040 1.0000