10/14/24, 11:32 AM A4 - Jupyter Notebook

In [ ]: Name: Jay Uddhav Shinde

Roll No : 70
BE IT

In [1]: import pandas as pd

import numpy as np
import tensorflow as tf
import matplotlib.pyplot as plt
import seaborn as sns
from sklearn.model_selection import train_test_split

from sklearn.preprocessing import StandardScaler

from sklearn.metrics import confusion_matrix, recall_score, accuracy_score, precision_score

RANDOM_SEED = 2021
TEST_PCT = 0.3
LABELS = ["Normal","Fraud"]

In [2]: dataset = pd.read_csv("creditcard.csv")

In [3]: #check for any null values

print("Any nulls in the dataset",dataset.isnull().values.any())
print('-------')
print("No. of unique labels",len(dataset['Class'].unique()))
print("Label values",dataset.Class.unique())

#0 is for normal credit card transcation

#1 is for fraudulent credit card transcation
print('-------')
print("Break down of Normal and Fraud Transcations")
print(pd.value_counts(dataset['Class'],sort=True))

Any nulls in the dataset False

-------
No. of unique labels 2
Label values [0 1]
-------
Break down of Normal and Fraud Transcations
0 284315
1 492
Name: Class, dtype: int64

In [4]: #visualizing the imbalanced dataset

count_classes = pd.value_counts(dataset['Class'],sort=True)
count_classes.plot(kind='bar',rot=0)
plt.xticks(range(len(dataset['Class'].unique())),dataset.Class.unique())
plt.title("Frequency by observation number")
plt.xlabel("Class")
plt.ylabel("Number of Observations")

normal_dataset = dataset[dataset.Class == 0]
fraud_dataset = dataset[dataset.Class == 1]

localhost:8888/notebooks/Downloads/A4.ipynb 1/8
10/14/24, 11:32 AM A4 - Jupyter Notebook

In [5]: #Save the normal and fradulent transcations in seperate dataframe

normal_dataset = dataset[dataset.Class == 0] fraud_dataset =
dataset[dataset.Class == 1]

#Visualize transcation amounts for normal and fraudulent transcations bins =

np.linspace(200,2500,100)
plt.hist(normal_dataset.Amount,bins=bins,alpha=1,density=True,label='Normal')
plt.hist(fraud_dataset.Amount,bins=bins,alpha=0.5,density=True,label='Fraud')
plt.legend(loc='upper right') plt.title("Transcation Amount vs Percentage of
Transcaions") plt.xlabel("Transcation Amount (USD)") plt.ylabel("Percentage of
Transcations") plt.show()

In [6]: dataset
Out[6]:
Time V1 V2 V3 V4 V5 V6 V7 V8 V9 ... V21 V22 V23 V24
0.0 -1.359807 -0.072781 2.536347 1.378155 -0.338321 0.462388 0.239599 0.098698 0.363787 ... -0.018307 0.277838 -0.110474 0.066928
0
0.0 1.191857 0.266151 0.166480 0.448154 0.060018 -0.082361 -0.078803 0.085102 -0.255425 ... -0.225775 -0.638672 0.101288 -0.339846
1
1.0 -1.358354 -1.340163 1.773209 0.379780 -0.503198 1.800499 0.791461 0.247676 -1.514654 ... 0.247998 0.771679 0.909412 -0.68928
2
1.0 -0.966272 -0.185226 1.792993 -0.863291 -0.010309 1.247203 0.237609 0.377436 -1.387024 ... -0.108300 0.005274 -0.190321 -1.175575
3
2.0 -1.158233 0.877737 1.548718 0.403034 -0.407193 0.095921 0.592941 -0.270533 0.817739 ... -0.009431 0.798278 -0.137458 0.141267
4
... ... ... ... ... ... ... ... ... ... ... ... ... ...
... .
172786.0 -11.881118 10.071785 -9.834783 -2.066656 -5.364473 -2.606837 -4.918215 7.305334 1.914428 ... 0.213454 0.111864 1.014480
284802 -0.509348
172787.0 -0.732789 -0.055080 2.035030 -0.738589 0.868229 1.058415 0.024330 0.294869 0.584800 ... 0.214205 0.924384 0.012463 -1.016226
284803
172788.0 1.919565 -0.301254 -3.249640 -0.557828 2.630515 3.031260 -0.296827 0.708417 0.432454 ... 0.232045 0.578229 -0.037501
284804 0.640134
172788.0 -0.240440 0.530483 0.702510 0.689799 -0.377961 0.623708 -0.686180 0.679145 0.392087 ... 0.265245 0.800049 -0.163298 0.123205
284805
172792.0 -0.533413 -0.189733 0.703337 -0.506271 -0.012546 -0.649617 1.577006 -0.414650 0.486180 ... 0.261057 0.643078 0.376777 0.008797
284806

284807 rows × 31 columns

In [7]: sc = StandardScaler()
dataset['Time'] = sc.fit_transform(dataset['Time'].values.reshape(-1,1))
dataset['Amount'] = sc.fit_transform(dataset['Amount'].values.reshape(-1,1))

In [8]: raw_data = dataset.values

#The last element contains if the transcation is normal which is represented by 0 and if fraud then 1
labels = raw_data[:,-1]

#The other data points are the electrocadriogram data

data = raw_data[:,0:-1]

train_data,test_data,train_labels,test_labels = train_test_split(data,labels,test_size = 0.2,random_state =2021)

localhost:8888/notebooks/Downloads/A4.ipynb 2/8
10/14/24, 11:32 AM A4 - Jupyter Notebook

In [9]: min_val = tf.reduce_min(train_data) max_val =

tf.reduce_max(train_data)

train_data = (train_data - min_val) / (max_val - min_val)

test_data = (test_data - min_val) / (max_val - min_val)

train_data = tf.cast(train_data,tf.float32) test_data =

tf.cast(test_data,tf.float32)

In [10]: train_labels = train_labels.astype(bool)

test_labels = test_labels.astype(bool)

#Creating normal and fraud datasets

normal_train_data = train_data[~train_labels]
normal_test_data = test_data[~test_labels]

fraud_train_data = train_data[train_labels]
fraud_test_data = test_data[test_labels]
print("No. of records in Fraud Train Data=",len(fraud_train_data))
print("No. of records in Normal Train Data=",len(normal_train_data))
print("No. of records in Fraud Test Data=",len(fraud_test_data))
print("No. of records in Normal Test Data=",len(normal_test_data))
No. of records in Fraud Train Data= 389
No. of records in Normal Train Data= 227456
No. of records in Fraud Test Data= 103
No. of records in Normal Test Data= 56859

In [11]: nb_epoch = 50
batch_size = 64
input_dim = normal_train_data.shape[1]
#num of columns,30
encoding_dim = 14
hidden_dim1 = int(encoding_dim / 2)
hidden_dim2 = 4
learning_rate = 1e-7

localhost:8888/notebooks/Downloads/A4.ipynb 3/8
10/14/24, 11:32 AM A4 - Jupyter Notebook

In [12]: #input layer

# Build models
input_layer = tf.keras.layers.Input(shape=(input_dim,))

#Encoder
encoder = tf.keras.layers.Dense(encoding_dim,activation="tanh",activity_regularizer = tf.keras.regularizers.l2(learning_rate))(in
encoder = tf.keras.layers.Dropout(0.2)(encoder)
encoder = tf.keras.layers.Dense(hidden_dim1,activation='relu')(encoder)
encoder = tf.keras.layers.Dense(hidden_dim2,activation=tf.nn.leaky_relu)(encoder)

#Decoder
decoder = tf.keras.layers.Dense(hidden_dim1,activation='relu')(encoder)
decoder = tf.keras.layers.Dropout(0.2)(decoder)
decoder = tf.keras.layers.Dense(encoding_dim,activation='relu')(decoder)
decoder = tf.keras.layers.Dense(input_dim,activation='tanh')(decoder)

#Autoencoder
# combine encoder and decoder to create the autoencoder model
autoencoder = tf.keras.Model(inputs = input_layer,outputs = decoder)
autoencoder.summary()

Model: "model"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
input_1 (InputLayer) [(None, 30)] 0

dense (Dense) (None, 14) 434

dropout (Dropout) (None, 14) 0

dense_1 (Dense) (None, 7) 105

dense_2 (Dense) (None, 4) 32

dense_3 (Dense) (None, 7) 35

dropout_1 (Dropout) (None, 7) 0

dense_4 (Dense) (None, 14) 112

dense_5 (Dense) (None, 30) 450

=================================================================
Total params: 1168 (4.56 KB)
Trainable params: 1168 (4.56 KB)
Non-trainable params: 0 (0.00 Byte)
_________________________________________________________________

In [13]: cp = tf.keras.callbacks.ModelCheckpoint(filepath="autoencoder_fraud.h5",mode='min',monitor='val_loss',verbose=2,save_best_only=Tr
#Define our early stopping
early_stop = tf.keras.callbacks.EarlyStopping(
monitor='val_loss',
min_delta=0.0001,
patience=10, verbose=11,
mode='min',
restore_best_weights=True

In [14]: autoencoder.compile(metrics=['accuracy'],loss= 'mean_squared_error',optimizer='adam')

localhost:8888/notebooks/Downloads/A4.ipynb 4/8
10/14/24, 11:32 AM A4 - Jupyter Notebook

In [15]: history = autoencoder.fit(normal_train_data,normal_train_data,epochs = nb_epoch,

batch_size = batch_size,shuffle = True,
validation_data = (test_data,test_data),
verbose=1,
callbacks = [cp,early_stop]).history
Epoch 1/50
3538/3554 [============================>.] - ETA: 0s - loss: 0.0046 - accuracy: 0.0455
Epoch 1: val_loss improved from inf to 0.00002, saving model to autoencoder_fraud.h5
3554/3554 [==============================] - 6s 2ms/step - loss: 0.0046 - accuracy: 0.0457 - val_loss: 2.0015e-05 - val_accurac
y: 0.0814
Epoch 2/50
81/3554 [..............................] - ETA: 5s - loss: 1.8520e-05 - accuracy: 0.0538

C:\Users\ASUS\AppData\Roaming\Python\Python39\site-packages\keras\src\engine\training.py:3000: UserWarning: You are saving your

model as an HDF5 file via `model.save()`. This file format is considered legacy. We recommend using instead the native Keras fo
rmat, e.g. `model.save('my_model.keras')`.
saving_api.save_model(

3529/3554 [============================>.] - ETA: 0s - loss: 1.9470e-05 - accuracy: 0.0694

Epoch 2: val_loss did not improve from 0.00002
3554/3554 [==============================] - 6s 2ms/step - loss: 1.9465e-05 - accuracy: 0.0694 - val_loss: 2.0077e-05 - val_acc
uracy: 0.0556
Epoch 3/50
3551/3554 [============================>.] - ETA: 0s - loss: 1.9567e-05 - accuracy: 0.0657
Epoch 3: val_loss did not improve from 0.00002
3554/3554 [==============================] - 6s 2ms/step - loss: 1.9568e-05 - accuracy: 0.0657 - val_loss: 2.0048e-05 - val_acc
uracy: 0.0251
Epoch 4/50
3518/3554 [============================>.] - ETA: 0s - loss: 1.9552e-05 - accuracy: 0.0578
Epoch 4: val_loss did not improve from 0.00002
3554/3554 [==============================] - 6s 2ms/step - loss: 1.9549e-05 - accuracy: 0.0576 - val_loss: 2.0063e-05 - val_acc
uracy: 0.0420
Epoch 5/50
3548/3554 [============================>.] - ETA: 0s - loss: 1.9589e-05 - accuracy: 0.0624
Epoch 5: val_loss did not improve from 0.00002
3554/3554 [==============================] - 6s 2ms/step - loss: 1.9583e-05 - accuracy: 0.0624 - val_loss: 2.0040e-05 - val_acc
uracy: 0.0371
Epoch 6/50
3550/3554 [============================>.] - ETA: 0s - loss: 1.9535e-05 - accuracy: 0.0586
Epoch 6: val_loss did not improve from 0.00002
3554/3554 [==============================] - 5s 2ms/step - loss: 1.9535e-05 - accuracy: 0.0585 - val_loss: 2.0027e-05 - val_acc
uracy: 0.0111
Epoch 7/50
3544/3554 [============================>.] - ETA: 0s - loss: 1.9531e-05 - accuracy: 0.0637
Epoch 7: val_loss did not improve from 0.00002
3554/3554 [==============================] - 6s 2ms/step - loss: 1.9530e-05 - accuracy: 0.0637 - val_loss: 2.0045e-05 - val_acc
uracy: 0.0371
Epoch 8/50
3537/3554 [============================>.] - ETA: 0s - loss: 1.9527e-05 - accuracy: 0.0600
Epoch 8: val_loss did not improve from 0.00002
3554/3554 [==============================] - 5s 1ms/step - loss: 1.9522e-05 - accuracy: 0.0602 - val_loss: 2.0126e-05 - val_acc
uracy: 6.1444e-04
Epoch 9/50
3529/3554 [============================>.] - ETA: 0s - loss: 1.9514e-05 - accuracy: 0.0591
Epoch 9: val_loss did not improve from 0.00002
3554/3554 [==============================] - 6s 2ms/step - loss: 1.9509e-05 - accuracy: 0.0590 - val_loss: 2.0123e-05 - val_acc
uracy: 0.0251
Epoch 10/50
3546/3554 [============================>.] - ETA: 0s - loss: 1.9510e-05 - accuracy: 0.0610
Epoch 10: val_loss did not improve from 0.00002
3554/3554 [==============================] - 6s 2ms/step - loss: 1.9512e-05 - accuracy: 0.0609 - val_loss: 2.0510e-05 - val_acc
uracy: 0.0111
Epoch 11/50
3531/3554 [============================>.] - ETA: 0s - loss: 1.9504e-05 - accuracy: 0.0614
Epoch 11: val_loss did not improve from 0.00002
Restoring model weights from the end of the best epoch: 1.
3554/3554 [==============================] - 6s 2ms/step - loss: 1.9498e-05 - accuracy: 0.0614 - val_loss: 2.0291e-05 - val_acc
uracy: 0.0814
Epoch 11: early stopping

localhost:8888/notebooks/Downloads/A4.ipynb 5/8
10/14/24, 11:32 AM A4 - Jupyter Notebook

In [16]: plt.plot(history['loss'],linewidth = 2,label = 'Train')

plt.plot(history['val_loss'],linewidth = 2,label = 'Test')
plt.legend(loc='upper right')
plt.title('Model Loss')
plt.ylabel('Loss')
plt.xlabel('Epoch')

#plt.ylim(ymin=0.70,ymax=1)

plt.show()

In [17]: test_x_predictions = autoencoder.predict(test_data)

mse = np.mean(np.power(test_data - test_x_predictions, 2),axis = 1)
error_df = pd.DataFrame({'Reconstruction_error':mse,
'True_class':test_labels})

1781/1781 [==============================] - 2s 950us/step

localhost:8888/notebooks/Downloads/A4.ipynb 6/8
10/14/24, 11:32 AM A4 - Jupyter Notebook

In [18]: threshold_fixed = 50
groups = error_df.groupby('True_class')
fig,ax = plt.subplots()

for name,group in groups:

ax.plot(group.index,group.Reconstruction_error,marker='o',ms = 3.5,linestyle='',
label = "Fraud" if name==1 else "Normal")
ax.hlines(threshold_fixed,ax.get_xlim()[0],ax.get_xlim()[1],colors="r",zorder=100,label="Threshold")
ax.legend()
plt.title("Reconstructions error for normal and fraud data")
plt.ylabel("Reconstruction error")
plt.xlabel("Data point index")
plt.show()

localhost:8888/notebooks/Downloads/A4.ipynb 7/8
10/14/24, 11:32 AM A4 - Jupyter Notebook

In [19]: threshold_fixed = 52
pred_y = [1 if e > threshold_fixed else 0
for e in
error_df.Reconstruction_error.values]
error_df['pred'] = pred_y
conf_matrix = confusion_matrix(error_df.True_class,pred_y)

plt.figure(figsize = (4,4))
sns.heatmap(conf_matrix,xticklabels = LABELS,yticklabels = LABELS,annot = True,fmt="d")
plt.title("Confusion matrix")
plt.ylabel("True class")
plt.xlabel("Predicted class")
plt.show()

#Print Accuracy,Precision and Recall

print("Accuracy :",accuracy_score(error_df['True_class'],error_df['pred']))
print("Recall :",recall_score(error_df['True_class'],error_df['pred']))
print("Precision :",precision_score(error_df['True_class'],error_df['pred']))

Accuracy : 0.9981917769741231
Recall : 0.0
Precision : 0.0
C:\ProgramData\Anaconda3\lib\site-packages\sklearn\metrics\_classification.py:1318: UndefinedMetricWarning: Precision is ill-de
fined and being set to 0.0 due to no predicted samples. Use `zero_division` parameter to control this behavior.
_warn_prf(average, modifier, msg_start, len(result))

In [ ]:

localhost:8888/notebooks/Downloads/A4.ipynb 8/8