# 1. Import library
import pandas as pd
import numpy as np

from sklearn.preprocessing import LabelEncoder

# 2. Baca dataset asli (yang bersih, bukan _noise)
df = pd.read_csv('dataset_klasifikasi_modul9_clean.csv')

# 3. Kalau datamu masih 1 kolom panjang, gunakan split ini.
#    Kalau sudah 7 kolom, bagian ini boleh dilewati.
if df.shape[1] == 1:
    df = df.iloc[:, 0].str.replace('"', '', regex=False).str.split(',', expand=True)
    df.columns = ['CustomerID','Usia','Pendapatan','FrekuensiBelanja','TotalBelanja','TrenPenjualan','Respon_Promo']

# Handle the header row issue and convert to numeric types for the main df
df = df.iloc[1:] # Drop the header row if it's still present as data
numeric_cols = ['Usia','Pendapatan','FrekuensiBelanja','TotalBelanja','Respon_Promo']
for col in numeric_cols:
    df[col] = pd.to_numeric(df[col], errors='coerce')
df.dropna(inplace=True)

# 4. Encode kolom kategori TrenPenjualan
encoder = LabelEncoder()
df['TrenPenjualan_encoded'] = encoder.fit_transform(df['TrenPenjualan'])

# 5. BUAT DATAFRAME NOISE DI SINI
df_noisy = df.copy()   # <-- INI YANG PENTING, MEMBUAT df_noisy

# 5a. Noise numerik pada Pendapatan (±5%)
df_noisy['Pendapatan'] = (
    df_noisy['Pendapatan'] *
    (1 + np.random.uniform(-0.05, 0.05, size=len(df_noisy)))
).astype(int)

# 5b. Noise numerik pada TotalBelanja (±10%)
df_noisy['TotalBelanja'] = (
    df_noisy['TotalBelanja'] *
    (1 + np.random.uniform(-0.10, 0.10, size=len(df_noisy)))
).astype(int)

# 5c. Noise label Respon_Promo (10% baris di-flip)
n_noise = int(0.10 * len(df_noisy))
idx_noise = np.random.choice(df_noisy.index, size=n_noise, replace=False)
df_noisy.loc[idx_noise, 'Respon_Promo'] = 1 - df_noisy.loc[idx_noise, 'Respon_Promo']

# 6. Cek sekilas
print(df_noisy.head())
print(df_noisy['Respon_Promo'].value_counts(normalize=True))