Support Vector Machine (SVM)
সম্পূর্ণ বাংলা লেকচার নোট

# প্রয়োজনীয় লাইব্রেরি
import pandas as pd
import numpy as np
from sklearn.svm import SVC
from sklearn.preprocessing import StandardScaler
from sklearn.model_selection import train_test_split
from sklearn.metrics import accuracy_score, confusion_matrix, classification_report
import matplotlib.pyplot as plt

# Dataset
data = {
    'study_hours':    [1,2,2,3,4,4,5,5,6,6,7,7,8,8,9,3,5,7,2,8],
    'attendance':     [40,50,45,55,60,70,65,75,80,85,90,88,95,92,98,48,72,88,42,96],
    'previous_marks': [30,40,35,45,50,55,60,58,65,70,75,72,80,78,88,38,62,79,33,84],
    'pass_fail':      [0,0,0,0,0,1,0,1,1,1,1,1,1,1,1,0,1,1,0,1]
}
df = pd.DataFrame(data)
print(df.head(8))

X = df[['study_hours', 'attendance', 'previous_marks']]
y = df['pass_fail']

# Train-Test Split আগে
X_train, X_test, y_train, y_test = train_test_split(
    X, y, test_size=0.2, random_state=42
)

# তারপর Scaling
scaler = StandardScaler()
X_train_sc = scaler.fit_transform(X_train)
X_test_sc  = scaler.transform(X_test)

print("Scaling সম্পন্ন!")
print("আগে (study_hours range):", X_train['study_hours'].min(), "to", X_train['study_hours'].max())
print("পরে (scaled range):", X_train_sc[:,0].min().round(2), "to", X_train_sc[:,0].max().round(2))

# Linear SVM
svm_linear = SVC(kernel='linear', C=1.0, random_state=42)
svm_linear.fit(X_train_sc, y_train)

y_pred_l = svm_linear.predict(X_test_sc)
print("✅ Linear SVM Accuracy:", accuracy_score(y_test, y_pred_l))
print("\nSupport Vectors সংখ্যা:", svm_linear.n_support_)
print("(ফেল শ্রেণীর, পাস শ্রেণীর)")
print("\n📊 Confusion Matrix:")
print(confusion_matrix(y_test, y_pred_l))

# RBF Kernel SVM
svm_rbf = SVC(kernel='rbf', C=1.0, gamma='scale', random_state=42)
svm_rbf.fit(X_train_sc, y_train)

y_pred_r = svm_rbf.predict(X_test_sc)
print("✅ RBF SVM Accuracy:", accuracy_score(y_test, y_pred_r))
print("\n📋 Classification Report:")
print(classification_report(y_test, y_pred_r, target_names=['ফেল', 'পাস']))

# Scaling ছাড়া SVM
svm_no_scale = SVC(kernel='rbf', random_state=42)
svm_no_scale.fit(X_train, y_train)
acc_no_scale = accuracy_score(y_test, svm_no_scale.predict(X_test))

# Scaling সহ SVM (আগে দেখা)
acc_scaled = accuracy_score(y_test, y_pred_r)

print(f"Scaling ছাড়া SVM Accuracy: {acc_no_scale:.2%}")
print(f"Scaling সহ SVM Accuracy:  {acc_scaled:.2%}")
print("\n💡 পার্থক্য স্পষ্ট! Feature Scaling SVM-এ বাধ্যতামূলক।")

print("C মান  | Train Accuracy | Test Accuracy")
print("-" * 42)
for c in [0.01, 0.1, 1, 10, 100]:
    m = SVC(kernel='rbf', C=c, gamma='scale', random_state=42)
    m.fit(X_train_sc, y_train)
    tr = accuracy_score(y_train, m.predict(X_train_sc))
    te = accuracy_score(y_test,  m.predict(X_test_sc))
    print(f"C={c:6} | {tr:.2%}         | {te:.2%}")