Spaces:
Sleeping
Sleeping
Added RF folder
Browse files
RF/__pycache__/grappaBroskiSkibiddiLewandoski.cpython-312.pyc
ADDED
|
Binary file (3.14 kB). View file
|
|
|
RF/__pycache__/grappaBroskiSkibiddiLewandoski.cpython-313.pyc
ADDED
|
Binary file (3.2 kB). View file
|
|
|
RF/__pycache__/test.cpython-312.pyc
ADDED
|
Binary file (2.47 kB). View file
|
|
|
RF/__pycache__/test.cpython-313.pyc
ADDED
|
Binary file (2.59 kB). View file
|
|
|
RF/grappaBroskiSkibiddiLewandoski.py
ADDED
|
@@ -0,0 +1,45 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import numpy as np
|
| 2 |
+
from DT.DecisionTree import DecisionTree
|
| 3 |
+
|
| 4 |
+
class HobaTitoMambo:
|
| 5 |
+
|
| 6 |
+
def __init__(self, n_trees=10, max_depth=10, max_features="sqrt", min_samples_split=2, min_sample_leafs=1, criterion="gini",random_state=None):
|
| 7 |
+
|
| 8 |
+
self.n_trees = n_trees
|
| 9 |
+
self.max_depth = max_depth
|
| 10 |
+
self.max_features = max_features
|
| 11 |
+
self.min_samples_split = min_samples_split
|
| 12 |
+
self.min_sample_leafs = min_sample_leafs
|
| 13 |
+
self.criterion = criterion
|
| 14 |
+
self.random_state = random_state
|
| 15 |
+
self.trees = []
|
| 16 |
+
self.rng = np.random.default_rng(self.random_state)
|
| 17 |
+
|
| 18 |
+
def __bootstrap_samples(self, x, y):
|
| 19 |
+
n_samples = x.shape[0]
|
| 20 |
+
idxs = self.rng.choice(n_samples, size=n_samples, replace=True)
|
| 21 |
+
return x[idxs], y[idxs]
|
| 22 |
+
|
| 23 |
+
def fit(self, x, y):
|
| 24 |
+
self.trees = []
|
| 25 |
+
|
| 26 |
+
for i in range(self.n_trees):
|
| 27 |
+
tree_seed = self.random_state + i if self.random_state is not None else None
|
| 28 |
+
|
| 29 |
+
dt = DecisionTree(maxDepth=self.max_depth,minSamplesSplit=self.min_samples_split,criterion=self.criterion,maxFeatures=self.max_features,minSampleLeafs=self.min_sample_leafs,randomState=tree_seed)
|
| 30 |
+
|
| 31 |
+
x_s, y_s = self.__bootstrap_samples(x, y)
|
| 32 |
+
dt.fit(x_s, y_s)
|
| 33 |
+
self.trees.append(dt)
|
| 34 |
+
|
| 35 |
+
def predict(self, x_test):
|
| 36 |
+
preds = np.array([tree.predict(x_test) for tree in self.trees])
|
| 37 |
+
preds = preds.T
|
| 38 |
+
|
| 39 |
+
res = np.zeros(len(x_test), dtype=int)
|
| 40 |
+
|
| 41 |
+
for i, sample_votes in enumerate(preds):
|
| 42 |
+
vals, counts = np.unique(sample_votes, return_counts=True)
|
| 43 |
+
res[i] = vals[np.argmax(counts)]
|
| 44 |
+
|
| 45 |
+
return res
|
RF/test.py
ADDED
|
@@ -0,0 +1,70 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import numpy as np
|
| 2 |
+
import pickle
|
| 3 |
+
from tensorflow import keras
|
| 4 |
+
# from sklearn.metrics import accuracy_score, f1_score, classification_report, confusion_matrix
|
| 5 |
+
from RF.grappaBroskiSkibiddiLewandoski import HobaTitoMambo
|
| 6 |
+
from Metrics import classification_report2, confusion_matrix2, accuracy_score2
|
| 7 |
+
|
| 8 |
+
#Download Data
|
| 9 |
+
(X_train, y_train), (X_test, y_test) = keras.datasets.mnist.load_data()
|
| 10 |
+
|
| 11 |
+
X_train = X_train / 255.0
|
| 12 |
+
X_test = X_test / 255.0
|
| 13 |
+
|
| 14 |
+
|
| 15 |
+
from skimage.feature import hog
|
| 16 |
+
|
| 17 |
+
def extract_hog_features(X):
|
| 18 |
+
features = []
|
| 19 |
+
|
| 20 |
+
for img in X:
|
| 21 |
+
img_2d = img.reshape(28, 28)
|
| 22 |
+
|
| 23 |
+
hog_features = hog(
|
| 24 |
+
img_2d,
|
| 25 |
+
orientations=9,
|
| 26 |
+
pixels_per_cell=(4, 4),
|
| 27 |
+
cells_per_block=(2, 2),
|
| 28 |
+
block_norm='L2-Hys',
|
| 29 |
+
feature_vector=True
|
| 30 |
+
)
|
| 31 |
+
|
| 32 |
+
features.append(hog_features)
|
| 33 |
+
|
| 34 |
+
return np.array(features)
|
| 35 |
+
|
| 36 |
+
|
| 37 |
+
X_train_HOG = extract_hog_features(X_train)
|
| 38 |
+
X_test_HOG = extract_hog_features(X_test)
|
| 39 |
+
|
| 40 |
+
print("Initializing HobaTitoMambo Random Forest...")
|
| 41 |
+
rf = HobaTitoMambo(
|
| 42 |
+
n_trees=10,
|
| 43 |
+
max_depth=15,
|
| 44 |
+
min_samples_split=20,
|
| 45 |
+
min_sample_leafs=5,
|
| 46 |
+
max_features="sqrt",
|
| 47 |
+
criterion="gini",
|
| 48 |
+
random_state=42
|
| 49 |
+
)
|
| 50 |
+
|
| 51 |
+
print("Training model (this may take a few minutes)...")
|
| 52 |
+
rf.fit(X_train_HOG, y_train)
|
| 53 |
+
|
| 54 |
+
print("Predicting on test set...")
|
| 55 |
+
predictions = rf.predict(X_test_HOG)
|
| 56 |
+
|
| 57 |
+
conf_matrix = confusion_matrix2(y_test, predictions)
|
| 58 |
+
|
| 59 |
+
print("\n--- Evaluation Results ---")
|
| 60 |
+
print("\nConfusion Matrix:")
|
| 61 |
+
print(conf_matrix)
|
| 62 |
+
|
| 63 |
+
print("\nDetailed Classification Report:")
|
| 64 |
+
print(classification_report2(y_test, predictions))
|
| 65 |
+
|
| 66 |
+
model_filename = 'rf_hog_model.pkl'
|
| 67 |
+
print(f"\nSaving model to {model_filename}...")
|
| 68 |
+
with open(model_filename, 'wb') as file:
|
| 69 |
+
pickle.dump(rf, file)
|
| 70 |
+
print("Model saved successfully!")
|