Create mymodel.py
Browse filesadapter le code pour la librairie FHE
- mymodel.py +170 -0
mymodel.py
ADDED
@@ -0,0 +1,170 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# Import Dependencies
|
2 |
+
import yaml
|
3 |
+
from joblib import dump, load
|
4 |
+
import pandas as pd
|
5 |
+
from sklearn.model_selection import train_test_split, cross_val_score
|
6 |
+
from sklearn.metrics import accuracy_score, confusion_matrix, classification_report
|
7 |
+
# Naive Bayes Approach
|
8 |
+
from sklearn.naive_bayes import MultinomialNB
|
9 |
+
# Trees Approach
|
10 |
+
from sklearn.tree import DecisionTreeClassifier
|
11 |
+
# Ensemble Approach
|
12 |
+
from sklearn.ensemble import RandomForestClassifier, GradientBoostingClassifier
|
13 |
+
import seaborn as sn
|
14 |
+
import matplotlib.pyplot as plt
|
15 |
+
|
16 |
+
|
17 |
+
class DiseasePrediction:
|
18 |
+
# Initialize and Load the Config File
|
19 |
+
def __init__(self, model_name=None):
|
20 |
+
# Load Config File
|
21 |
+
try:
|
22 |
+
with open('./config.yaml', 'r') as f:
|
23 |
+
self.config = yaml.safe_load(f)
|
24 |
+
except Exception as e:
|
25 |
+
print("Error reading Config file...")
|
26 |
+
|
27 |
+
# Verbose
|
28 |
+
self.verbose = self.config['verbose']
|
29 |
+
# Load Training Data
|
30 |
+
self.train_features, self.train_labels, self.train_df = self._load_train_dataset()
|
31 |
+
# Load Test Data
|
32 |
+
self.test_features, self.test_labels, self.test_df = self._load_test_dataset()
|
33 |
+
# Feature Correlation in Training Data
|
34 |
+
self._feature_correlation(data_frame=self.train_df, show_fig=False)
|
35 |
+
# Model Definition
|
36 |
+
self.model_name = model_name
|
37 |
+
# Model Save Path
|
38 |
+
self.model_save_path = self.config['model_save_path']
|
39 |
+
|
40 |
+
# Function to Load Train Dataset
|
41 |
+
def _load_train_dataset(self):
|
42 |
+
df_train = pd.read_csv(self.config['dataset']['training_data_path'])
|
43 |
+
cols = df_train.columns
|
44 |
+
cols = cols[:-2]
|
45 |
+
train_features = df_train[cols]
|
46 |
+
train_labels = df_train['prognosis']
|
47 |
+
|
48 |
+
# Check for data sanity
|
49 |
+
assert (len(train_features.iloc[0]) == 132)
|
50 |
+
assert (len(train_labels) == train_features.shape[0])
|
51 |
+
|
52 |
+
if self.verbose:
|
53 |
+
print("Length of Training Data: ", df_train.shape)
|
54 |
+
print("Training Features: ", train_features.shape)
|
55 |
+
print("Training Labels: ", train_labels.shape)
|
56 |
+
return train_features, train_labels, df_train
|
57 |
+
|
58 |
+
# Function to Load Test Dataset
|
59 |
+
def _load_test_dataset(self):
|
60 |
+
df_test = pd.read_csv(self.config['dataset']['test_data_path'])
|
61 |
+
cols = df_test.columns
|
62 |
+
cols = cols[:-1]
|
63 |
+
test_features = df_test[cols]
|
64 |
+
test_labels = df_test['prognosis']
|
65 |
+
|
66 |
+
# Check for data sanity
|
67 |
+
assert (len(test_features.iloc[0]) == 132)
|
68 |
+
assert (len(test_labels) == test_features.shape[0])
|
69 |
+
|
70 |
+
if self.verbose:
|
71 |
+
print("Length of Test Data: ", df_test.shape)
|
72 |
+
print("Test Features: ", test_features.shape)
|
73 |
+
print("Test Labels: ", test_labels.shape)
|
74 |
+
return test_features, test_labels, df_test
|
75 |
+
|
76 |
+
# Features Correlation
|
77 |
+
def _feature_correlation(self, data_frame=None, show_fig=False):
|
78 |
+
# Get Feature Correlation
|
79 |
+
corr = data_frame.corr()
|
80 |
+
sn.heatmap(corr, square=True, annot=False, cmap="YlGnBu")
|
81 |
+
plt.title("Feature Correlation")
|
82 |
+
plt.tight_layout()
|
83 |
+
if show_fig:
|
84 |
+
plt.show()
|
85 |
+
plt.savefig('feature_correlation.png')
|
86 |
+
|
87 |
+
# Dataset Train Validation Split
|
88 |
+
def _train_val_split(self):
|
89 |
+
X_train, X_val, y_train, y_val = train_test_split(self.train_features, self.train_labels,
|
90 |
+
test_size=self.config['dataset']['validation_size'],
|
91 |
+
random_state=self.config['random_state'])
|
92 |
+
|
93 |
+
if self.verbose:
|
94 |
+
print("Number of Training Features: {0}\tNumber of Training Labels: {1}".format(len(X_train), len(y_train)))
|
95 |
+
print("Number of Validation Features: {0}\tNumber of Validation Labels: {1}".format(len(X_val), len(y_val)))
|
96 |
+
return X_train, y_train, X_val, y_val
|
97 |
+
|
98 |
+
# Model Selection
|
99 |
+
def select_model(self):
|
100 |
+
if self.model_name == 'mnb':
|
101 |
+
self.clf = MultinomialNB()
|
102 |
+
elif self.model_name == 'decision_tree':
|
103 |
+
self.clf = DecisionTreeClassifier(criterion=self.config['model']['decision_tree']['criterion'])
|
104 |
+
elif self.model_name == 'random_forest':
|
105 |
+
self.clf = RandomForestClassifier(n_estimators=self.config['model']['random_forest']['n_estimators'])
|
106 |
+
elif self.model_name == 'gradient_boost':
|
107 |
+
self.clf = GradientBoostingClassifier(n_estimators=self.config['model']['gradient_boost']['n_estimators'],
|
108 |
+
criterion=self.config['model']['gradient_boost']['criterion'])
|
109 |
+
return self.clf
|
110 |
+
|
111 |
+
# ML Model
|
112 |
+
def train_model(self):
|
113 |
+
# Get the Data
|
114 |
+
X_train, y_train, X_val, y_val = self._train_val_split()
|
115 |
+
classifier = self.select_model()
|
116 |
+
# Training the Model
|
117 |
+
classifier = classifier.fit(X_train, y_train)
|
118 |
+
# Trained Model Evaluation on Validation Dataset
|
119 |
+
confidence = classifier.score(X_val, y_val)
|
120 |
+
# Validation Data Prediction
|
121 |
+
y_pred = classifier.predict(X_val)
|
122 |
+
# Model Validation Accuracy
|
123 |
+
accuracy = accuracy_score(y_val, y_pred)
|
124 |
+
# Model Confusion Matrix
|
125 |
+
conf_mat = confusion_matrix(y_val, y_pred)
|
126 |
+
# Model Classification Report
|
127 |
+
clf_report = classification_report(y_val, y_pred)
|
128 |
+
# Model Cross Validation Score
|
129 |
+
score = cross_val_score(classifier, X_val, y_val, cv=3)
|
130 |
+
|
131 |
+
if self.verbose:
|
132 |
+
print('\nTraining Accuracy: ', confidence)
|
133 |
+
print('\nValidation Prediction: ', y_pred)
|
134 |
+
print('\nValidation Accuracy: ', accuracy)
|
135 |
+
print('\nValidation Confusion Matrix: \n', conf_mat)
|
136 |
+
print('\nCross Validation Score: \n', score)
|
137 |
+
print('\nClassification Report: \n', clf_report)
|
138 |
+
|
139 |
+
# Save Trained Model
|
140 |
+
dump(classifier, str(self.model_save_path + self.model_name + ".joblib"))
|
141 |
+
|
142 |
+
# Function to Make Predictions on Test Data
|
143 |
+
def make_prediction(self, saved_model_name=None, test_data=None):
|
144 |
+
try:
|
145 |
+
# Load Trained Model
|
146 |
+
clf = load(str(self.model_save_path + saved_model_name + ".joblib"))
|
147 |
+
except Exception as e:
|
148 |
+
print("Model not found...")
|
149 |
+
|
150 |
+
if test_data is not None:
|
151 |
+
result = clf.predict(test_data)
|
152 |
+
return result
|
153 |
+
else:
|
154 |
+
result = clf.predict(self.test_features)
|
155 |
+
accuracy = accuracy_score(self.test_labels, result)
|
156 |
+
clf_report = classification_report(self.test_labels, result)
|
157 |
+
return accuracy, clf_report
|
158 |
+
|
159 |
+
|
160 |
+
if __name__ == "__main__":
|
161 |
+
# Model Currently Training
|
162 |
+
current_model_name = 'decision_tree'
|
163 |
+
# Instantiate the Class
|
164 |
+
dp = DiseasePrediction(model_name=current_model_name)
|
165 |
+
# Train the Model
|
166 |
+
dp.train_model()
|
167 |
+
# Get Model Performance on Test Data
|
168 |
+
test_accuracy, classification_report = dp.make_prediction(saved_model_name=current_model_name)
|
169 |
+
print("Model Test Accuracy: ", test_accuracy)
|
170 |
+
print("Test Data Classification Report: \n", classification_report)
|