knitr::opts_chunk$set(echo = TRUE, warning = FALSE, message = FALSE)
library(readxl)
library(dplyr)
## Warning: package 'dplyr' was built under R version 4.6.2
## 
## Attaching package: 'dplyr'
## The following objects are masked from 'package:stats':
## 
##     filter, lag
## The following objects are masked from 'package:base':
## 
##     intersect, setdiff, setequal, union
library(tidyr)
## Warning: package 'tidyr' was built under R version 4.6.2
library(tseries)
## Warning: package 'tseries' was built under R version 4.6.2
## Registered S3 method overwritten by 'quantmod':
##   method            from
##   as.zoo.data.frame zoo
library(forecast)
## Warning: package 'forecast' was built under R version 4.6.2
library(ggplot2)
## Warning: package 'ggplot2' was built under R version 4.6.2
file_path <- "C:/Users/User/OneDrive/Documents/data 15 negara asia.xlsx"

data_raw <- tryCatch(
  readxl::read_excel(file_path, col_names = FALSE),
  error = function(e) {
    read.delim(file_path, header = FALSE, stringsAsFactors = FALSE)
  }
)
## New names:
## • `` -> `...1`
## • `` -> `...2`
## • `` -> `...3`
## • `` -> `...4`
## • `` -> `...5`
## • `` -> `...6`
## • `` -> `...7`
## • `` -> `...8`
## • `` -> `...9`
## • `` -> `...10`
## • `` -> `...11`
## • `` -> `...12`
## • `` -> `...13`
## • `` -> `...14`
## • `` -> `...15`
## • `` -> `...16`
## • `` -> `...17`
## • `` -> `...18`
## • `` -> `...19`
## • `` -> `...20`
## • `` -> `...21`
## • `` -> `...22`
## • `` -> `...23`
## • `` -> `...24`
## • `` -> `...25`
## • `` -> `...26`
## • `` -> `...27`
## • `` -> `...28`
bersihkan_seri <- function(row_vec) {
  vals <- as.character(unlist(row_vec))
  vals <- vals[!grepl("Philipine|Philippines|Pertumbuhan Ekonomi|akses Listrik|Penguna Internet|Pengguna Telepon|Seluler", vals, ignore.case = TRUE)]
  vals <- gsub("\u00A0", "", vals, fixed = TRUE)
  vals <- trimws(vals)
  vals[vals %in% c("", "..", "...", "-", "NA", "na", "null")] <- NA

  has_comma <- grepl(",", vals, fixed = TRUE)
  has_dot   <- grepl(".", vals, fixed = TRUE)
  vals[has_comma & !has_dot] <- gsub(",", ".", vals[has_comma & !has_dot], fixed = TRUE)
  vals[has_comma & has_dot]  <- gsub(",", "", vals[has_comma & has_dot], fixed = TRUE)

  suppressWarnings(as.numeric(vals))
}

idx_matches <- which(apply(data_raw, 1, function(r) {
  any(grepl("Pertumbuhan Ekonomi", as.character(r), ignore.case = TRUE))
}))
stopifnot("Baris 'Pertumbuhan Ekonomi' tidak ditemukan" = length(idx_matches) > 0)

idx_y <- NULL
y_num <- numeric(0)
for (cand in idx_matches) {
  tmp_y <- bersihkan_seri(data_raw[cand, ])
  if (any(!is.na(tmp_y))) {
    idx_y <- cand
    y_num <- tmp_y
    break
  }
}
stopifnot("Semua nilai Y bernilai NA; cek struktur tabel" = length(y_num) > 0 && any(!is.na(y_num)))

x1_num <- bersihkan_seri(data_raw[min(idx_y + 1, nrow(data_raw)), ])
x2_num <- bersihkan_seri(data_raw[min(idx_y + 2, nrow(data_raw)), ])
x3_num <- bersihkan_seri(data_raw[min(idx_y + 3, nrow(data_raw)), ])

n_valid <- max(which(!is.na(y_num)))
y_num  <- y_num[1:min(n_valid, length(y_num))]
x1_num <- x1_num[1:min(n_valid, length(x1_num))]
x2_num <- x2_num[1:min(n_valid, length(x2_num))]
x3_num <- x3_num[1:min(n_valid, length(x3_num))]

ts_y  <- ts(y_num, start = 1, frequency = 1)
ts_x1 <- ts(x1_num, start = 1, frequency = 1)
ts_x2 <- ts(x2_num, start = 1, frequency = 1)
ts_x3 <- ts(x3_num, start = 1, frequency = 1)

# Interpolasi nilai NA internal agar deret waktu kontinu untuk pemodelan
ts_y_interp  <- na.interp(ts_y)
ts_x1_interp <- na.interp(ts_x1)
ts_x2_interp <- na.interp(ts_x2)
ts_x3_interp <- na.interp(ts_x3)

summary(ts_y_interp)
##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##  -2.066   4.705   5.033   4.657   5.529   6.345
autoplot(ts_y_interp) +
  ggtitle("Pertumbuhan Ekonomi (Y) - Philipine") +
  xlab("Periode") + ylab("Pertumbuhan Ekonomi (Y)") +
  theme_minimal()

adf_y <- adf.test(as.numeric(ts_y_interp))
print(adf_y)
## 
##  Augmented Dickey-Fuller Test
## 
## data:  as.numeric(ts_y_interp)
## Dickey-Fuller = -2.8141, Lag order = 2, p-value = 0.2617
## alternative hypothesis: stationary
ts_y_diff <- diff(ts_y_interp, differences = 1)
adf_y_diff <- adf.test(as.numeric(ts_y_diff))
print(adf_y_diff)
## 
##  Augmented Dickey-Fuller Test
## 
## data:  as.numeric(ts_y_diff)
## Dickey-Fuller = -3.8715, Lag order = 2, p-value = 0.03061
## alternative hypothesis: stationary
par(mfrow = c(1, 2))
acf(as.numeric(ts_y_diff), main = "ACF - Diff Y")
pacf(as.numeric(ts_y_diff), main = "PACF - Diff Y")

par(mfrow = c(1, 1))
seri_stasioner <- if (adf_y$p.value < 0.05) ts_y_interp else ts_y_diff

model_ar1 <- Arima(seri_stasioner, order = c(1, 0, 0))
summary(model_ar1)
## Series: seri_stasioner 
## ARIMA(1,0,0) with non-zero mean 
## 
## Coefficients:
##           ar1    mean
##       -0.3051  0.1481
## s.e.   0.1834  0.2770
## 
## sigma^2 = 3.613:  log likelihood = -52.6
## AIC=111.2   AICc=112.29   BIC=114.97
## 
## Training set error measures:
##                      ME     RMSE      MAE      MPE     MAPE      MASE
## Training set 0.01162016 1.826185 1.002611 125.2095 205.1218 0.6591956
##                     ACF1
## Training set -0.09093682
model_ma1 <- Arima(seri_stasioner, order = c(0, 0, 1))
summary(model_ma1)
## Series: seri_stasioner 
## ARIMA(0,0,1) with non-zero mean 
## 
## Coefficients:
##           ma1    mean
##       -0.5374  0.1105
## s.e.   0.2026  0.1695
## 
## sigma^2 = 3.287:  log likelihood = -51.49
## AIC=108.98   AICc=110.08   BIC=112.76
## 
## Training set error measures:
##                     ME     RMSE      MAE       MPE     MAPE      MASE
## Training set 0.0688841 1.741919 1.040498 -41.21833 285.1407 0.6841058
##                    ACF1
## Training set 0.07104386
model_arma11 <- Arima(seri_stasioner, order = c(1, 0, 1))
summary(model_arma11)
## Series: seri_stasioner 
## ARIMA(1,0,1) with non-zero mean 
## 
## Coefficients:
##          ar1      ma1    mean
##       0.4464  -1.0000  0.0374
## s.e.  0.2045   0.1537  0.0696
## 
## sigma^2 = 3.077:  log likelihood = -51.11
## AIC=110.22   AICc=112.12   BIC=115.25
## 
## Training set error measures:
##                    ME     RMSE      MAE       MPE     MAPE      MASE       ACF1
## Training set 0.303052 1.649938 0.945872 0.1942829 182.5838 0.6218909 -0.0313797
model_auto_y <- auto.arima(ts_y_interp, stepwise = FALSE, approximation = FALSE)
summary(model_auto_y)
## Series: ts_y_interp 
## ARIMA(1,0,0) with non-zero mean 
## 
## Coefficients:
##          ar1    mean
##       0.3751  4.5892
## s.e.  0.1916  0.4930
## 
## sigma^2 = 2.857:  log likelihood = -51.52
## AIC=109.04   AICc=110.09   BIC=112.93
## 
## Training set error measures:
##                      ME     RMSE       MAE      MPE     MAPE      MASE
## Training set 0.05956235 1.626532 0.9228585 7.193199 35.76399 0.9355407
##                     ACF1
## Training set -0.03940746
xreg_mat <- cbind(
  X1 = as.numeric(ts_x1_interp),
  X2 = as.numeric(ts_x2_interp),
  X3 = as.numeric(ts_x3_interp)
)

model_arimax <- auto.arima(ts_y_interp, xreg = xreg_mat, stepwise = FALSE, approximation = FALSE)
summary(model_arimax)
## Series: ts_y_interp 
## Regression with ARIMA(0,0,0) errors 
## 
## Coefficients:
##       intercept      X1       X2       X3
##        -48.8033  0.6122  -0.0596  -0.0317
## s.e.    20.9917  0.2419   0.0210   0.0191
## 
## sigma^2 = 2.639:  log likelihood = -49.25
## AIC=108.49   AICc=111.35   BIC=114.97
## 
## Training set error measures:
##                        ME     RMSE       MAE      MPE     MAPE      MASE
## Training set -4.68876e-13 1.499303 0.9543634 2.885731 33.72993 0.9674785
##                   ACF1
## Training set 0.1255779
model_x1 <- auto.arima(ts_x1_interp, stepwise = FALSE, approximation = FALSE)
model_x2 <- auto.arima(ts_x2_interp, stepwise = FALSE, approximation = FALSE)
model_x3 <- auto.arima(ts_x3_interp, stepwise = FALSE, approximation = FALSE)
tabel_aic <- data.frame(
  Model = c("AR(1)", "MA(1)", "ARMA(1,1)", "Auto ARIMA (Y)", "ARIMAX (Y ~ X1+X2+X3)"),
  AIC = c(
    AIC(model_ar1),
    AIC(model_ma1),
    AIC(model_arma11),
    AIC(model_auto_y),
    AIC(model_arimax)
  )
)
print(tabel_aic)
##                   Model      AIC
## 1                 AR(1) 111.1984
## 2                 MA(1) 108.9850
## 3             ARMA(1,1) 110.2166
## 4        Auto ARIMA (Y) 109.0426
## 5 ARIMAX (Y ~ X1+X2+X3) 108.4927
checkresiduals(model_auto_y)

## 
##  Ljung-Box test
## 
## data:  Residuals from ARIMA(1,0,0) with non-zero mean
## Q* = 0.2124, df = 4, p-value = 0.9947
## 
## Model df: 1.   Total lags used: 5
ramalan_y <- forecast(model_auto_y, h = 5)
print(ramalan_y)
##    Point Forecast    Lo 80    Hi 80    Lo 95    Hi 95
## 28       4.783821 2.617561 6.950080 1.470813 8.096829
## 29       4.662202 2.348595 6.975809 1.123846 8.200559
## 30       4.616589 2.283002 6.950177 1.047675 8.185503
## 31       4.599482 2.263097 6.935866 1.026290 8.172673
## 32       4.593065 2.256288 6.929843 1.019273 8.166858
autoplot(ramalan_y) +
  ggtitle("Forecasting Pertumbuhan Ekonomi (Y) - Philipine") +
  xlab("Periode") + ylab("Pertumbuhan Ekonomi (Y)") +
  theme_minimal()