df <- read_excel("Dataset_EDA_30_Observasi.xlsx", sheet = "Data") %>%
  mutate(Tanggal = as.Date(Tanggal))

head(df, 10)

dim(df)

names(df)

str(df)

tabel_tipe <- tibble(
  Variabel = names(df),
  Jenis_data = c("Temporal", "Kualitatif", "Kualitatif",
                 "Kuantitatif diskrit", "Kuantitatif kontinu", "Kuantitatif kontinu"),
  Skala_pengukuran = c("Interval", "Nominal", "Nominal", "Rasio", "Rasio", "Rasio"),
  Alasan = c(
    "Tanggal memiliki jarak sama antarhari tetapi tidak memiliki nol mutlak.",
    "Online dan Offline merupakan kategori tanpa tingkatan.",
    "Reguler, Express, dan Premium merupakan kategori layanan (ubah ke ordinal jika kamu menganggap berurutan, beri alasan).",
    "Jumlah kejadian transaksi berupa hitungan dan memiliki nol mutlak.",
    "Durasi proses dapat memiliki nilai pecahan dan nol berarti tidak ada durasi.",
    "Nilai uang memiliki nol mutlak dan dapat dibandingkan secara rasio."
  )
)
tampil(tabel_tipe, caption = "Identifikasi jenis data dan skala pengukuran", booktabs = TRUE)

cek_kualitas <- tibble(
  Pemeriksaan = c("Jumlah baris", "Jumlah kolom", "Missing value",
                  "Baris duplikat", "Tanggal paling awal", "Tanggal paling akhir"),
  Hasil = c(
    nrow(df), ncol(df),
    sum(is.na(df)),
    sum(duplicated(df)),
    as.character(min(df$Tanggal, na.rm = TRUE)),
    as.character(max(df$Tanggal, na.rm = TRUE))
  )
)
tampil(cek_kualitas, caption = "Ringkasan pemeriksaan kualitas data")

missing_kolom <- tibble(
  Variabel = names(df),
  Jumlah_missing = sapply(df, function(x) sum(is.na(x)))
)
tampil(missing_kolom, caption = "Missing value pada setiap variabel")

buat_frekuensi <- function(x) {
  hasil <- as.data.frame(table(x), stringsAsFactors = FALSE)
  names(hasil) <- c("Kategori", "Frekuensi")
  hasil %>%
    mutate(
      Frekuensi_Relatif = Frekuensi / sum(Frekuensi),
      Persentase = percent(Frekuensi_Relatif, accuracy = 0.1),
      Frekuensi_Kumulatif = cumsum(Frekuensi)
    )
}

freq_kanal <- buat_frekuensi(df$Kanal)
freq_layanan <- buat_frekuensi(df$Jenis_Layanan)

tampil(freq_kanal, caption = "Distribusi frekuensi Kanal")

tampil(freq_layanan, caption = "Distribusi frekuensi Jenis Layanan")

modus <- function(x) {
  x <- x[!is.na(x)]
  nilai <- unique(x)
  frek <- tabulate(match(x, nilai))
  nilai[frek == max(frek)]
}

variabel_numerik <- c("Jumlah_Transaksi", "Waktu_Proses_Menit",
                      "Nilai_Transaksi_Juta_Rp")

stat_deskriptif <- map_dfr(variabel_numerik, function(v) {
  x <- df[[v]]
  tibble(
    Variabel = v,
    Mean = mean(x, na.rm = TRUE),
    Median = median(x, na.rm = TRUE),
    Modus = paste(modus(x), collapse = ", "),
    Minimum = min(x, na.rm = TRUE),
    Maksimum = max(x, na.rm = TRUE),
    Rentang = diff(range(x, na.rm = TRUE)),
    Varians = var(x, na.rm = TRUE),
    Standar_Deviasi = sd(x, na.rm = TRUE),
    CV_persen = sd(x, na.rm = TRUE) / mean(x, na.rm = TRUE) * 100
  )
})
tampil(stat_deskriptif, digits = 2,
      caption = "Statistik deskriptif variabel numerik")

# Jumlah kelas Sturges
k_sturges <- ceiling(1 + 3.322 * log10(nrow(df)))
k_sturges

# Pembanding mean-median dan urutan CV
stat_deskriptif %>%
  mutate(Mean_minus_Median = Mean - Median) %>%
  select(Variabel, Mean, Median, Mean_minus_Median, CV_persen) %>%
  arrange(desc(CV_persen)) %>%
  tampil(digits = 2, caption = "Pembanding mean-median dan koefisien variasi")

ggplot(df, aes(x = Kanal, fill = Kanal)) +
  geom_bar(width = 0.65, show.legend = FALSE) +
  geom_text(stat = "count", aes(label = after_stat(count)), vjust = -0.5) +
  scale_fill_manual(values = c("Online" = "#4C91B8", "Offline" = "#E6A15C")) +
  labs(title = "Frekuensi Transaksi Berdasarkan Kanal",
       x = "Kanal transaksi", y = "Jumlah pengamatan") +
  tema_uts

pie_data <- df %>% count(Jenis_Layanan) %>%
  mutate(label = paste0(Jenis_Layanan, "\n", percent(n / sum(n), accuracy = 0.1)))
ggplot(pie_data, aes(x = "", y = n, fill = Jenis_Layanan)) +
  geom_col(width = 1, color = "white") +
  coord_polar(theta = "y") +
  geom_text(aes(label = label), position = position_stack(vjust = 0.5),
            size = 3.5) +
  labs(title = "Proporsi Jenis Layanan", x = NULL, y = NULL, fill = "Layanan") +
  tema_uts + theme(axis.text = element_blank(), axis.ticks = element_blank(),
                   panel.grid = element_blank())

par(mfrow = c(3, 1), mar = c(4, 4, 3, 1))
for (v in variabel_numerik) {
  x <- df[[v]]
  hist(x, breaks = "Sturges", col = "#8DB7CC", border = "white",
       main = paste("Histogram", v),
       xlab = v, ylab = "Frekuensi")
}

par(mfrow = c(1, 1))

data_long <- df %>%
  select(all_of(variabel_numerik)) %>%
  pivot_longer(everything(), names_to = "Variabel", values_to = "Nilai")

ggplot(data_long, aes(x = "", y = Nilai)) +
  geom_boxplot(fill = "#8DB7CC", color = "#24536A", width = 0.45) +
  facet_wrap(~ Variabel, scales = "free_y", ncol = 1) +
  labs(title = "Boxplot Variabel Numerik", x = NULL, y = "Nilai") +
  tema_uts

df %>%
  mutate(Obs_ke = row_number()) %>%
  pivot_longer(all_of(variabel_numerik), names_to = "Variabel", values_to = "Nilai") %>%
  group_by(Variabel) %>%
  filter(Nilai == max(Nilai) | Nilai == min(Nilai)) %>%
  arrange(Variabel, Nilai) %>%
  tampil(digits = 2, caption = "Nilai minimum dan maksimum tiap variabel")

ringkasan_kanal <- df %>%
  group_by(Kanal) %>%
  summarise(
    Jumlah_hari = n(),
    Rata_rata_transaksi = mean(Jumlah_Transaksi, na.rm = TRUE),
    Median_transaksi = median(Jumlah_Transaksi, na.rm = TRUE),
    Rata_rata_waktu = mean(Waktu_Proses_Menit, na.rm = TRUE),
    Rata_rata_nilai = mean(Nilai_Transaksi_Juta_Rp, na.rm = TRUE),
    .groups = "drop"
  )
tampil(ringkasan_kanal, digits = 2,
      caption = "Perbandingan ringkasan berdasarkan kanal")

ringkasan_kanal_lengkap <- df %>%
  group_by(Kanal) %>%
  summarise(across(all_of(variabel_numerik),
                   list(Mean = mean, Median = median, SD = sd),
                   .names = "{.col}__{.fn}"), .groups = "drop") %>%
  pivot_longer(-Kanal, names_to = c("Variabel", "Statistik"),
               names_sep = "__", values_to = "Nilai") %>%
  pivot_wider(names_from = Statistik, values_from = Nilai)

tampil(ringkasan_kanal_lengkap, digits = 2,
      caption = "Mean, median, dan SD ketiga variabel berdasarkan kanal")

ggplot(ringkasan_kanal, aes(x = Kanal, y = Rata_rata_transaksi, fill = Kanal)) +
  geom_col(width = 0.6, show.legend = FALSE) +
  geom_text(aes(label = round(Rata_rata_transaksi, 1)), vjust = -0.5) +
  scale_fill_manual(values = c("Online" = "#4C91B8", "Offline" = "#E6A15C")) +
  labs(title = "Rata-rata Jumlah Transaksi Berdasarkan Kanal",
       x = "Kanal", y = "Rata-rata transaksi per hari") +
  tema_uts

ggplot(df, aes(x = Kanal, y = Waktu_Proses_Menit, fill = Kanal)) +
  geom_boxplot(alpha = 0.85, show.legend = FALSE) +
  scale_fill_manual(values = c("Online" = "#4C91B8", "Offline" = "#E6A15C")) +
  labs(title = "Sebaran Waktu Proses Berdasarkan Kanal",
       x = "Kanal", y = "Waktu proses (menit)") +
  tema_uts

ggplot(df, aes(x = Jenis_Layanan, y = Nilai_Transaksi_Juta_Rp,
               fill = Jenis_Layanan)) +
  geom_boxplot(show.legend = FALSE, alpha = 0.85) +
  scale_fill_manual(values = c("Reguler" = "#8DB7CC",
                               "Express" = "#E6A15C",
                               "Premium" = "#8BB99B")) +
  labs(title = "Sebaran Nilai Transaksi Berdasarkan Jenis Layanan",
       x = "Jenis layanan", y = "Nilai transaksi (juta rupiah)") +
  tema_uts

ggplot(df, aes(x = Jumlah_Transaksi, y = Nilai_Transaksi_Juta_Rp,
               color = Jenis_Layanan)) +
  geom_point(size = 2.8, alpha = 0.85) +
  geom_smooth(aes(group = 1), method = "lm", se = TRUE,
              color = "#333333", linewidth = 0.8) +
  labs(title = "Hubungan Jumlah Transaksi dan Nilai Transaksi",
       x = "Jumlah transaksi", y = "Nilai transaksi (juta rupiah)",
       color = "Jenis layanan") +
  tema_uts

korelasi <- cor(df$Jumlah_Transaksi, df$Nilai_Transaksi_Juta_Rp,
                use = "complete.obs")
tibble(Hubungan = "Jumlah transaksi dan nilai transaksi",
       Korelasi_Pearson = korelasi)

ggplot(df, aes(x = Jumlah_Transaksi, y = Waktu_Proses_Menit)) +
  geom_point(aes(color = Kanal), size = 2.8, alpha = 0.85) +
  geom_smooth(method = "lm", se = TRUE, color = "#333333", linewidth = 0.8) +
  labs(title = "Hubungan Jumlah Transaksi dan Waktu Proses",
       x = "Jumlah transaksi", y = "Waktu proses (menit)", color = "Kanal") +
  tema_uts

cor_tw <- cor(df$Jumlah_Transaksi, df$Waktu_Proses_Menit)
cor_tw

# Observasi yang menyimpang dari garis tren (|residual standar| > 2)
m_tw <- lm(Waktu_Proses_Menit ~ Jumlah_Transaksi, data = df)
m_tn <- lm(Nilai_Transaksi_Juta_Rp ~ Jumlah_Transaksi, data = df)

menyimpang <- bind_rows(
  df %>% mutate(Obs_ke = row_number(), Hubungan = "Transaksi vs Waktu",
                Residual_std = rstandard(m_tw)),
  df %>% mutate(Obs_ke = row_number(), Hubungan = "Transaksi vs Nilai",
                Residual_std = rstandard(m_tn))
) %>%
  filter(abs(Residual_std) > 2) %>%
  select(Hubungan, Obs_ke, Tanggal, Kanal, Jenis_Layanan, Residual_std)

tampil(menyimpang, digits = 2, caption = "Observasi menyimpang dari pola hubungan")

ggplot(df, aes(x = Jumlah_Transaksi, y = Nilai_Transaksi_Juta_Rp,
               color = Kanal, size = Waktu_Proses_Menit)) +
  geom_point(alpha = 0.7) +
  scale_color_manual(values = c("Online" = "#4C91B8", "Offline" = "#E6A15C")) +
  labs(title = "Jumlah vs Nilai Transaksi menurut Kanal dan Waktu Proses",
       x = "Jumlah transaksi", y = "Nilai transaksi (juta rupiah)",
       color = "Kanal", size = "Waktu proses (menit)") +
  tema_uts

df %>%
  group_by(Kanal) %>%
  summarise(Korelasi = cor(Jumlah_Transaksi, Nilai_Transaksi_Juta_Rp), .groups = "drop") %>%
  tampil(digits = 3, caption = "Korelasi jumlah dan nilai transaksi per kanal")

ggplot(df, aes(x = Jenis_Layanan, y = Waktu_Proses_Menit, fill = Kanal)) +
  geom_boxplot(position = position_dodge(width = 0.8), alpha = 0.85) +
  scale_fill_manual(values = c("Online" = "#4C91B8", "Offline" = "#E6A15C")) +
  labs(title = "Waktu Proses menurut Jenis Layanan dan Kanal",
       x = "Jenis layanan", y = "Waktu proses (menit)", fill = "Kanal") +
  tema_uts

count(df, Kanal, Jenis_Layanan) %>%
  tampil(caption = "Jumlah observasi per kombinasi kanal dan layanan")

tren_long <- df %>%
  select(Tanggal, Jumlah_Transaksi, Waktu_Proses_Menit) %>%
  pivot_longer(-Tanggal, names_to = "Variabel", values_to = "Nilai")

ggplot(tren_long, aes(x = Tanggal, y = Nilai, color = Variabel)) +
  geom_line(linewidth = 0.8) +
  geom_point(size = 1.8) +
  geom_smooth(method = "lm", se = FALSE, linetype = "dashed", linewidth = 0.6) +
  facet_wrap(~ Variabel, scales = "free_y", ncol = 1) +
  scale_x_date(date_labels = "%d %b", date_breaks = "5 days") +
  labs(title = "Perubahan Jumlah Transaksi dan Waktu Proses",
       x = "Tanggal", y = "Nilai", color = "Variabel") +
  tema_uts + theme(axis.text.x = element_text(angle = 45, hjust = 1))

df_tren <- df %>% mutate(Hari_ke = as.numeric(Tanggal - min(Tanggal)) + 1)
m_trx <- lm(Jumlah_Transaksi ~ Hari_ke, data = df_tren)
m_wkt <- lm(Waktu_Proses_Menit ~ Hari_ke, data = df_tren)

tibble(
  Variabel = c("Jumlah_Transaksi", "Waktu_Proses_Menit"),
  Slope_per_hari = c(coef(m_trx)[2], coef(m_wkt)[2]),
  R2 = c(summary(m_trx)$r.squared, summary(m_wkt)$r.squared),
  p_value = c(summary(m_trx)$coefficients[2, 4], summary(m_wkt)$coefficients[2, 4])
) %>%
  tampil(digits = 4, caption = "Kecenderungan tren linear selama 30 hari")

mat_korelasi <- cor(df %>% select(all_of(variabel_numerik)),
                    use = "pairwise.complete.obs")
korelasi_long <- as.data.frame(as.table(mat_korelasi)) %>%
  rename(Variabel_1 = Var1, Variabel_2 = Var2, Korelasi = Freq)

ggplot(korelasi_long, aes(x = Variabel_1, y = Variabel_2, fill = Korelasi)) +
  geom_tile(color = "white") +
  geom_text(aes(label = round(Korelasi, 2)), size = 4) +
  scale_fill_gradient2(low = "#C76D6D", mid = "white", high = "#4C91B8",
                       midpoint = 0, limits = c(-1, 1)) +
  labs(title = "Matriks Korelasi Variabel Numerik",
       x = NULL, y = NULL, fill = "Korelasi") +
  tema_uts + theme(axis.text.x = element_text(angle = 25, hjust = 1))

korelasi_long %>%
  filter(as.character(Variabel_1) < as.character(Variabel_2)) %>%
  arrange(desc(abs(Korelasi))) %>%
  tampil(digits = 3, caption = "Pasangan variabel diurutkan dari korelasi terkuat")

ringkasan_kombinasi <- df %>%
  group_by(Kanal, Jenis_Layanan) %>%
  summarise(
    Jumlah_hari = n(),
    Rata_rata_transaksi = mean(Jumlah_Transaksi, na.rm = TRUE),
    Rata_rata_waktu = mean(Waktu_Proses_Menit, na.rm = TRUE),
    Rata_rata_nilai = mean(Nilai_Transaksi_Juta_Rp, na.rm = TRUE),
    .groups = "drop"
  )
tampil(ringkasan_kombinasi, digits = 2,
      caption = "Ringkasan gabungan kanal dan jenis layanan")

ggplot(ringkasan_kombinasi,
       aes(x = Jenis_Layanan, y = Kanal, fill = Rata_rata_transaksi)) +
  geom_tile(color = "white") +
  geom_text(aes(label = round(Rata_rata_transaksi, 1)), size = 4) +
  scale_fill_gradient(low = "#EAF2F5", high = "#286783") +
  labs(title = "Rata-rata Transaksi Menurut Kanal dan Layanan",
       x = "Jenis layanan", y = "Kanal",
       fill = "Rata-rata transaksi") +
  tema_uts

jumlah_outlier <- hasil_outlier %>%
  group_by(Variabel) %>%
  summarise(
    Jumlah_IQR = sum(Outlier_IQR, na.rm = TRUE),
    Jumlah_z_di_atas_2 = sum(Perlu_diperhatikan_z, na.rm = TRUE),
    Jumlah_z_di_atas_3 = sum(Outlier_ekstrem_z, na.rm = TRUE),
    .groups = "drop"
  )
tampil(jumlah_outlier, caption = "Jumlah penanda outlier berdasarkan metode")

ggplot(data_long, aes(x = "", y = Nilai)) +
  geom_boxplot(fill = "#E6AD65", color = "#805A2B", outlier.color = "#B33A3A") +
  facet_wrap(~ Variabel, scales = "free_y", ncol = 1) +
  labs(title = "Pemeriksaan Visual Nilai Ekstrem",
       x = NULL, y = "Nilai") +
  tema_uts

# Data asli tidak diubah; hanya membandingkan ringkasan dengan dan tanpa outlier IQR
sensitivitas <- map_dfr(variabel_numerik, function(v) {
  d <- hasil_outlier %>% filter(Variabel == v)
  tibble(Variabel = v,
         Mean_semua = mean(d$Nilai),
         Mean_tanpa_outlier = mean(d$Nilai[!d$Outlier_IQR]),
         SD_semua = sd(d$Nilai),
         SD_tanpa_outlier = sd(d$Nilai[!d$Outlier_IQR]))
})
tampil(sensitivitas, digits = 2,
      caption = "Perbandingan mean dan SD dengan dan tanpa outlier IQR")

df_tanpa <- df %>% filter(!(Tanggal %in% tgl_outlier))
c(korelasi_semua = cor(df$Jumlah_Transaksi, df$Nilai_Transaksi_Juta_Rp),
  korelasi_tanpa_outlier = cor(df_tanpa$Jumlah_Transaksi, df_tanpa$Nilai_Transaksi_Juta_Rp))

kanal_tertinggi <- ringkasan_kanal %>% slice_max(Rata_rata_transaksi, n = 1, with_ties = FALSE)
kanal_waktu_terendah <- ringkasan_kanal %>% slice_min(Rata_rata_waktu, n = 1, with_ties = FALSE)
layanan_nilai_tertinggi <- ringkasan_kombinasi %>%
  slice_max(Rata_rata_nilai, n = 1, with_ties = FALSE)

cat(
  "Kanal dengan rata-rata transaksi tertinggi adalah", kanal_tertinggi$Kanal,
  "dengan rata-rata", round(kanal_tertinggi$Rata_rata_transaksi, 2),
  "transaksi per hari.\n\n",
  "Kanal dengan rata-rata waktu proses terendah adalah", kanal_waktu_terendah$Kanal,
  "dengan rata-rata", round(kanal_waktu_terendah$Rata_rata_waktu, 2),
  "menit.\n\n",
  "Kombinasi kanal dan layanan dengan rata-rata nilai transaksi tertinggi adalah",
  paste(layanan_nilai_tertinggi$Kanal,
        layanan_nilai_tertinggi$Jenis_Layanan, sep = " - "),
  "dengan rata-rata", round(layanan_nilai_tertinggi$Rata_rata_nilai, 2),
  "juta rupiah."
)

kpi_kanal <- df %>%
  group_by(Kanal) %>%
  summarise(
    Hari = n(),
    Total_transaksi = sum(Jumlah_Transaksi),
    Rata_rata_transaksi = mean(Jumlah_Transaksi),
    Rata_rata_nilai = mean(Nilai_Transaksi_Juta_Rp),
    Rata_rata_waktu = mean(Waktu_Proses_Menit),
    CV_waktu_persen = sd(Waktu_Proses_Menit) / mean(Waktu_Proses_Menit) * 100,
    .groups = "drop"
  )
tampil(kpi_kanal, digits = 2, caption = "Indikator kinerja per kanal")