File size: 4,680 Bytes
a40d10a
 
 
 
 
3d97c90
a40d10a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
import keras
from sklearn.metrics import accuracy_score, confusion_matrix, classification_report
from MNISTclassification.DT.DecisionTree import DecisionTree

# ======================
# 1. LOAD DATA
# ======================
(X_train, y_train), (X_test, y_test) = keras.datasets.mnist.load_data()

# Flatten
X_train = X_train.reshape(X_train.shape[0], -1)
X_test = X_test.reshape(X_test.shape[0], -1)

# Normalize
X_train = X_train / 255.0
X_test = X_test / 255.0

# ======================
# 2. BINARY LABELS (6 vs NOT 6)
# ======================
y_train_bin = (y_train == 6).astype(int)
y_test_bin  = (y_test == 6).astype(int)

# ======================
# 3. VARIANCE FILTER
# ======================
def compute_pixel_variance(X):
    mean = np.mean(X, axis=0)
    return np.mean((X - mean) ** 2, axis=0)

def variance_threshold(X, threshold=1e-4):
    variances = compute_pixel_variance(X)
    mask = variances > threshold
    return X[:, mask], mask

X_train_clean, mask = variance_threshold(X_train)
X_test_clean = X_test[:, mask]

print("After variance filtering:", X_train_clean.shape)

# ======================
# 4. BALANCE DATA (UNDERSAMPLING)
# ======================
def balance_data(X, y):
    idx_6 = np.where(y == 1)[0]
    idx_not6 = np.where(y == 0)[0]

    np.random.shuffle(idx_not6)
    idx_not6 = idx_not6[:len(idx_6)]

    idx_balanced = np.concatenate([idx_6, idx_not6])
    np.random.shuffle(idx_balanced)

    return X[idx_balanced], y[idx_balanced]

X_train_bal, y_train_bal = balance_data(X_train_clean, y_train_bin)

print("Balanced shape:", X_train_bal.shape)
print("Class counts:", np.bincount(y_train_bal))

# ======================
# 5. PCA (FULL)
# ======================
def pca_full(X):
    mean = np.mean(X, axis=0)
    X_centered = X - mean

    cov_matrix = np.cov(X_centered, rowvar=False)

    eigenvalues, eigenvectors = np.linalg.eigh(cov_matrix)

    idx = np.argsort(eigenvalues)[::-1]
    eigenvectors = eigenvectors[:, idx]
    eigenvalues = eigenvalues[idx]

    return eigenvalues, eigenvectors, mean

def choose_k(eigenvalues, threshold=0.95):
    explained_variance_ratio = eigenvalues / np.sum(eigenvalues)
    cumulative = np.cumsum(explained_variance_ratio)

    k = np.argmax(cumulative >= threshold) + 1
    return k, cumulative

# IMPORTANT: PCA ON BALANCED TRAINING DATA
eigenvalues, eigenvectors, mean = pca_full(X_train_bal)
k, cumulative = choose_k(eigenvalues, threshold=0.95)

print(f"\nOptimal number of components: {k}")

# ======================
# 6. PROJECT DATA
# ======================
components = eigenvectors[:, :k]

X_train_pca = (X_train_bal - mean) @ components
X_test_pca  = (X_test_clean - mean) @ components

print("Shape after PCA (train):", X_train_pca.shape)
print("Shape after PCA (test):", X_test_pca.shape)

# ======================
# 7. DECISION TREE
# ======================
dt = DecisionTree(maxDepth=8, minSamplesSplit=10)
X_train_small = X_train_bal[:5000]
y_train_small = y_train_bal[:5000]
dt.fit(X_train_small, y_train_small)

# Predict
predictions = dt.predict(X_test_pca)

# ======================
# 8. EVALUATION
# ======================
accuracy = accuracy_score(y_test_bin, predictions)
print("\nAccuracy:", accuracy)

cm = confusion_matrix(y_test_bin, predictions)
print("\nConfusion Matrix:\n", cm)

plt.figure(figsize=(6, 5))
sns.heatmap(cm, annot=True, fmt='d', cmap='Blues')
plt.xlabel("Predicted")
plt.ylabel("Actual")
plt.title("Confusion Matrix (1 vs Not 1)")
plt.show()

print("\nClassification Report:\n")
print(classification_report(
    y_test_bin, predictions,
    target_names=["Not 1", "Is 1"]
))

# ======================
# 9. PCA SPACE VISUALIZATION
# ======================
plt.figure(figsize=(8, 6))
plt.scatter(X_train_pca[:, 0], X_train_pca[:, 1],
            c=y_train_bal, cmap='coolwarm', s=5)
plt.xlabel("PC1")
plt.ylabel("PC2")
plt.title("PCA Space (1 vs Not 1)")
plt.colorbar()
plt.show()

# ======================
# 10. EXPLAINED VARIANCE
# ======================
plt.plot(cumulative)
plt.axhline(y=0.95, color='r', linestyle='--')
plt.xlabel("Number of Components")
plt.ylabel("Cumulative Variance")
plt.title("Explained Variance Curve")
plt.show()

# Optional: train accuracy on balanced train set
train_preds = dt.predict(X_train_pca)
print("Train Accuracy:", accuracy_score(y_train_bal, train_preds))

# Count classes in test set
num_6 = np.sum(y_test_bin == 1)
num_not6 = np.sum(y_test_bin == 0)

print("\nTest set distribution:")
print("Is 6:", num_6)
print("Not 6:", num_not6)

ratio = num_not6 / num_6
print("\nRatio (Not 6 : 6) =", ratio)
print("Ratio (6 : Not 6) = 1 :", ratio)