# Load library
library(randomForest)
## Warning: package 'randomForest' was built under R version 4.5.2
## randomForest 4.7-1.2
## Type rfNews() to see new features/changes/bug fixes.
library(dplyr)
## Warning: package 'dplyr' was built under R version 4.5.2
## 
## Attaching package: 'dplyr'
## The following object is masked from 'package:randomForest':
## 
##     combine
## The following objects are masked from 'package:stats':
## 
##     filter, lag
## The following objects are masked from 'package:base':
## 
##     intersect, setdiff, setequal, union
library(lubridate)
## Warning: package 'lubridate' was built under R version 4.5.2
## 
## Attaching package: 'lubridate'
## The following objects are masked from 'package:base':
## 
##     date, intersect, setdiff, union
library(ggplot2)
## 
## Attaching package: 'ggplot2'
## The following object is masked from 'package:randomForest':
## 
##     margin
library(Metrics)
## Warning: package 'Metrics' was built under R version 4.5.3
data <- read.csv("data_mts.csv")
# Periksa beberapa baris pertama data
head(data)
##      tanggal penjualan  suhu harga pengunjung
## 1 2023-01-01    200.00 28.00 50.00     500.00
## 2 2023-01-02    206.56 28.05 50.02     531.72
## 3 2023-01-03    208.31 27.08 51.80     551.55
## 4 2023-01-04    204.23 29.19 52.01     511.91
## 5 2023-01-05    200.40 27.36 52.98     485.06
## 6 2023-01-06    193.51 27.37 54.07     466.07
str(data)
## 'data.frame':    1000 obs. of  5 variables:
##  $ tanggal   : chr  "2023-01-01" "2023-01-02" "2023-01-03" "2023-01-04" ...
##  $ penjualan : num  200 207 208 204 200 ...
##  $ suhu      : num  28 28.1 27.1 29.2 27.4 ...
##  $ harga     : num  50 50 51.8 52 53 ...
##  $ pengunjung: num  500 532 552 512 485 ...
# 3. Feature Engineering Sesuai Karakteristik Soal
data_clean <- data %>%
  mutate(tanggal = as.Date(tanggal)) %>%
  arrange(tanggal) %>%
  mutate(
    # [POLA 1: TREN NAIK] -> Tambahkan indeks waktu berurutan
    trend_index = row_number(),
    
    # [POLA 2: SIKLUS MINGGUAN (Jumat Paling Sepi)] -> Faktor nama hari
    nama_hari = factor(wday(tanggal, label = TRUE, abbr = FALSE, week_start = 1)),
    is_jumat  = ifelse(wday(tanggal, week_start = 1) == 5, 1, 0), # Flag khusus Jumat
    
    # [POLA 3: MUSIMAN TAHUNAN] -> Bulan dan hari ke-n dalam setahun
    bulan     = month(tanggal),
    hari_thn  = yday(tanggal),
    
    # [TIME SERIES AUTOREGRESSIVE] -> Penjualan kemarin & minggu lalu
    penjualan_lag1 = lag(penjualan, 1), # H-1
    penjualan_lag7 = lag(penjualan, 7)  # H-7 (siklus mingguan yang sama)
  ) %>%
  # Hapus baris NA akibat lag 7 hari pertama
  na.omit()
# 4. Pembagian Data Train dan Test (80% Train, 20% Test)
# Data Time Series dibagi berurutan (tidak di-shuffle/acak)
proporsi_train <- 0.8
n_train        <- floor(nrow(data_clean) * proporsi_train)

data_train <- data_clean[1:n_train, ]
data_test  <- data_clean[(n_train + 1):nrow(data_clean), ]

cat("Rentang Data Latih (Train):", as.character(min(data_train$tanggal)), "sampai", as.character(max(data_train$tanggal)), "\n")
## Rentang Data Latih (Train): 2023-01-08 sampai 2025-03-11
cat("Rentang Data Uji (Test)   :", as.character(min(data_test$tanggal)), "sampai", as.character(max(data_test$tanggal)), "\n")
## Rentang Data Uji (Test)   : 2025-03-12 sampai 2025-09-26
cat("Jumlah baris train:", nrow(data_train), "| Jumlah baris test:", nrow(data_test), "\n\n")
## Jumlah baris train: 794 | Jumlah baris test: 199
# 5. Formula Model: Variabel Multivariat + Fitur Pola Waktu
formula_rf <- penjualan ~ suhu + harga + pengunjung + 
                          trend_index +          # <-- Tambahkan ini (Pola Tren)
                          nama_hari + is_jumat + # <-- Tambahkan ini (Pola Jumat Sepi)
                          bulan + hari_thn +     # <-- Tambahkan hari_thn (Pola Musiman Tahunan)
                          penjualan_lag1 + penjualan_lag7
# 6. Melatih Model Random Forest
set.seed(42) # Menjaga reproduktibilitas hasil
model_rf <- randomForest(
  formula = formula_rf,
  data = data_train,
  ntree = 500,        # Banyak pohon keputusan
  mtry = 3,           # Banyak variabel acak tiap split node
  importance = TRUE
)
# Tampilkan ringkasan model
print(model_rf)
## 
## Call:
##  randomForest(formula = formula_rf, data = data_train, ntree = 500,      mtry = 3, importance = TRUE) 
##                Type of random forest: regression
##                      Number of trees: 500
## No. of variables tried at each split: 3
## 
##           Mean of squared residuals: 42.7152
##                     % Var explained: 84.04
# 7. Melakukan Prediksi pada Data Uji (Test Set)
prediksi_test <- predict(model_rf, newdata = data_test)
# 8. Evaluasi Akurasi Model
actual <- data_test$penjualan
rmse_score <- rmse(actual, prediksi_test)
mae_score  <- mae(actual, prediksi_test)
r2_score   <- 1 - (sum((actual - prediksi_test)^2) / sum((actual - mean(actual))^2))

cat("================ HASIL EVALUASI MODEL ================\n")
## ================ HASIL EVALUASI MODEL ================
cat(sprintf("RMSE (Root Mean Squared Error) : %.4f\n", rmse_score))
## RMSE (Root Mean Squared Error) : 8.5485
cat(sprintf("MAE  (Mean Absolute Error)      : %.4f\n", mae_score))
## MAE  (Mean Absolute Error)      : 6.5801
cat(sprintf("R-squared (R²)                  : %.4f (%.2f%%)\n", r2_score, r2_score * 100))
## R-squared (R²)                  : 0.6118 (61.18%)
cat("======================================================\n\n")
## ======================================================
# 9. Visualisasi Grafik: Nilai Aktual vs Nilai Prediksi
df_plot <- data_test %>%
  select(tanggal, penjualan) %>%
  mutate(prediksi = prediksi_test)

plot_pred <- ggplot(df_plot, aes(x = tanggal)) +
  geom_line(aes(y = penjualan, color = "Aktual"), linewidth = 0.9) +
  geom_line(aes(y = prediksi, color = "Prediksi RF"), linewidth = 0.9, linetype = "dashed") +
  labs(
    title = "Time Series Forecasting: Penjualan Aktual vs Prediksi Random Forest",
    subtitle = paste("Periode Data Test:", min(df_plot$tanggal), "s/d", max(df_plot$tanggal)),
    x = "Tanggal",
    y = "Jumlah Penjualan",
    color = "Keterangan"
  ) +
  scale_color_manual(values = c("Aktual" = "#1f77b4", "Prediksi RF" = "#d62728")) +
  theme_minimal(base_size = 12) +
  theme(legend.position = "top")

print(plot_pred)

# 10. Visualisasi Feature Importance (Variabel yang Paling Berpengaruh)
importance_df <- data.frame(
  Feature = rownames(importance(model_rf)),
  IncMSE  = importance(model_rf)[, "%IncMSE"]
)

plot_imp <- ggplot(importance_df, aes(x = reorder(Feature, IncMSE), y = IncMSE)) +
  geom_col(fill = "steelblue") +
  coord_flip() +
  labs(
    title = "Tingkat Kepentingan Variabel (Feature Importance)",
    x = "Variabel Prediktor",
    y = "% Kenaikan MSE jika Variabel Dihilangkan (%IncMSE)"
  ) +
  theme_minimal(base_size = 12)

print(plot_imp)

# Grafik Boxplot Rata-rata Penjualan per Hari (Sudah Diperbaiki)
ggplot(data_clean, aes(x = nama_hari, y = penjualan, fill = (nama_hari == "Friday"))) +
  geom_boxplot(alpha = 0.8, outlier.color = "red") +
  scale_fill_manual(values = c("FALSE" = "#6baed6", "TRUE" = "#e6550d"), 
                    labels = c("Hari Lain", "Jumat (Paling Sepi)"),
                    name = "Kategori Hari") +
  labs(
    title = "Verifikasi Pola Siklus Mingguan: Rata-rata Penjualan per Hari",
    subtitle = "Terbukti penjualan pada hari Jumat mengalami penurunan paling signifikan",
    x = "Hari",
    y = "Penjualan (unit/hari)"
  ) +
  theme_minimal(base_size = 12) +
  theme(legend.position = "top")