Studi Kasus Komprehensif
& Proyek Akhir

import pandas as pd
import numpy as np
from sklearn.model_selection import train_test_split, cross_val_score
from sklearn.preprocessing import StandardScaler
from sklearn.ensemble import RandomForestClassifier, RandomForestRegressor
from sklearn.linear_model import LinearRegression, LogisticRegression
from sklearn.metrics import r2_score, mean_absolute_error, classification_report
from sklearn.impute import SimpleImputer

np.random.seed(42)
n = 500

# ── Simulasi dataset mahasiswa ────────────────────
ipk_lalu   = np.random.normal(2.9, 0.55, n).clip(1.5, 4.0)
kehadiran  = (50 + 10*ipk_lalu + np.random.normal(0,8,n)).clip(40,100)
jam_belajar= (2  + 2.5*ipk_lalu + np.random.normal(0,2,n)).clip(1,20)
jam_hp     = (8  - jam_belajar + np.random.normal(0,1.5,n)).clip(1,12)
kerja      = np.random.binomial(1, 0.3, n)
prodi      = np.random.choice(['IF','SI','Math'], n)
nilai_akhir= (20 + 12*ipk_lalu + 0.5*kehadiran + 1.5*jam_belajar
              - 1.2*jam_hp - 3*kerja + np.random.normal(0,4,n)).clip(0,100)

df = pd.DataFrame({
    'ipk_lalu': ipk_lalu, 'kehadiran': kehadiran,
    'jam_belajar': jam_belajar, 'jam_hp': jam_hp,
    'kerja': kerja, 'prodi': prodi, 'nilai_akhir': nilai_akhir
})
# Tambahkan missing values buatan
df.loc[np.random.choice(df.index, 23, replace=False), 'kehadiran'] = np.nan

# ── Preprocessing ───────────────────────────────
df['kehadiran'] = df.groupby('prodi')['kehadiran'].transform(
    lambda x: x.fillna(x.median()))
df['rasio_belajar_hp'] = df.'jam_belajar'] / df['jam_hp']
prodi_dummies = pd.get_dummies(df['prodi'], prefix='prodi', drop_first=True)
df = pd.concat([df, prodi_dummies], axis=1)

features = ['ipk_lalu','kehadiran','jam_belajar','jam_hp',
            'kerja','rasio_belajar_hp','prodi_Math','prodi_SI']
X = df[features]
y_reg = df['nilai_akhir']
y_cls = (df['nilai_akhir'] >= 60).astype(int)

X_tr,X_te,yr_tr,yr_te = train_test_split(X, y_reg, test_size=0.2, random_state=42)
sc = StandardScaler()
X_trs = sc.fit_transform(X_tr); X_tes = sc.transform(X_te)

# ── Regresi ─────────────────────────────────────
rf_reg = RandomForestRegressor(200, random_state=42).fit(X_tr, yr_tr)
y_pred = rf_reg.predict(X_te)
print(f"Regresi — R²: {r2_score(yr_te,y_pred):.3f} | MAE: {mean_absolute_error(yr_te,y_pred):.2f}")

# Feature importance
imp = pd.Series(rf_reg.feature_importances_, index=features).sort_values(ascending=False)
print("\nFeature Importance:")
print(imp.round(3))

# ── Klasifikasi ──────────────────────────────────
_,X_te2,yc_tr,yc_te = train_test_split(X, y_cls, test_size=0.2, random_state=42)
rf_cls = RandomForestClassifier(200, random_state=42)
X_all_tr = X.iloc[X_tr.index]
rf_cls.fit(X_tr, y_cls.iloc[X_tr.index])
print("\nKlasifikasi:")
print(classification_report(yc_te, rf_cls.predict(X_te),
      target_names=['Tidak Lulus','Lulus']))

from sklearn.cluster import KMeans
from sklearn.decomposition import PCA
from sklearn.metrics import silhouette_score
from sklearn.preprocessing import StandardScaler
import pandas as pd; import numpy as np

np.random.seed(42)
n = 2300

# Simulasi 3 segmen nasabah bank
data = pd.DataFrame({
  'saldo_jt'      : np.concatenate([np.random.normal(85,20,644),
                     np.random.normal(12,5,805), np.random.normal(3.5,2,851)]).clip(0.5),
  'freq_trx'      : np.concatenate([np.random.normal(42,8,644),
                     np.random.normal(18,5,805), np.random.normal(4,2,851)]).clip(0),
  'recency_hari'   : np.concatenate([np.random.normal(3,1,644),
                     np.random.normal(12,5,805), np.random.normal(60,20,851)]).clip(1),
  'usia'          : np.concatenate([np.random.normal(48,8,644),
                     np.random.normal(28,4,805), np.random.normal(52,10,851)]).clip(18,75),
  'lama_tahun'    : np.concatenate([np.random.normal(9,3,644),
                     np.random.normal(1.5,0.8,805), np.random.normal(8,3,851)]).clip(0.5),
})

# Scaling wajib sebelum clustering
sc = StandardScaler()
X = sc.fit_transform(data)

# Elbow method + Silhouette untuk pilih K
print("K  | Inertia  | Silhouette")
for k in range(2, 7):
    km = KMeans(k, random_state=42, n_init=10).fit(X)
    sil = silhouette_score(X, km.labels_)
    print(f"{k}  | {km.inertia_:8.1f} | {sil:.3f}")

# Model terbaik K=3
km3 = KMeans(3, random_state=42, n_init=10)
data['cluster'] = km3.fit_predict(X)
print("\n=== PROFIL CLUSTER ===")
print(data.groupby('cluster').agg('mean').round(1))

# PCA 2D untuk visualisasi
pca = PCA(2)
X2d = pca.fit_transform(X)
print(f"\nPCA explained variance: {pca.explained_variance_ratio_.round(3)}")
print(f"Total: {pca.explained_variance_ratio_.sum():.1%} informasi tertangkap di 2D")