I'm new to machine learning world and currently in learning stage as a beginner. I have built and trained a model for fraud detection for a dataset of transactions held on a store.
Here's the complete code of my model:
df_full = pd.read_excel('input/invoiced_products_noinvoiceids_inproduct_v2.0.xlsx', sheet_name=0,)
df_full = df_full[df_full.filter(regex='^(?!Unnamed)').columns]
df_full.drop(['paymentdetails',], 1, inplace=True)
df_full.drop(['timestamp'], 1, inplace=True)
# Handle non numaric data
def handle_non_numaric_data(df_full):
columns = df_full.columns.values
for column in columns:
text_digit_vals = {}
def convert_to_int(val):
return text_digit_vals[val]
if df_full[column].dtype != np.int64 and df_full[column].dtype != np.float64:
column_contents = df_full[column].values.tolist()
unique_elements = set(column_contents)
x = 0
for unique in unique_elements:
if unique not in text_digit_vals:
text_digit_vals[unique] = x
x+=1
df_full[column] = list(map(convert_to_int, df_full[column]))
return df_full
df_full = handle_non_numaric_data(df_full)
print(df_full.head())
#for convert to numeric
df_full['discount'] = pd.to_numeric(df_full['discount'], errors='coerce')
df_full['productdiscount'] = pd.to_numeric(df_full['discount'], errors='coerce')
df_full['Class'] = ((df_full['discount'] > 20) &
(df_full['tax'] == 0) &
(df_full['productdiscount'] > 20) &
(df_full['total'] > 100)).astype(int)
# print (df_full)
df_full.to_csv('InvoiceData.csv')
# Get some sample data from entire dataset
data = df_full.sample(frac = 0.5, random_state = 1)
print(data.shape)
data.isnull().sum()
# Convert excel data into matrix
columns = "invoiceid locationid timestamp customerid discount tax total subtotal productid quantity productprice productdiscount invoice_products_id producttax invoice_payments_id paymentmethod paymentdetails amount Class(0/1) Class".split()
X = pd.DataFrame.as_matrix(data, columns=columns)
Y = data.Class
# temp = np.array(temp).reshape((len(temp), 1)
Y = Y.values.reshape(Y.shape[0], 1)
X.shape
X_train, X_test, Y_train, Y_test = train_test_split(X, Y, test_size = 0.06)
X_test, X_dev, Y_test, Y_dev = train_test_split(X_test, Y_test, test_size = .5)
# Check if there is Classification Values - 0/1 in training set and other set
np.where(Y_train == 1)
np.where(Y_test == 1)
np.where(Y_dev == 1)
# Determine no of fraud cases in dataset
Fraud = data[data['Class'] == 1]
Valid = data[data['Class'] == 0]
# calculate percentages for Fraud & Valid
outlier_fraction = len(Fraud) / float(len(Valid))
print('Fraud Cases : {}'.format(len(Fraud)))
print('Valid Cases : {}'.format(len(Valid)))
print(outlier_fraction)
# Get all the columns from dataframe
columns = data.columns.tolist()
# Filter the columns to remove data we don't want
columns = [c for c in columns if c not in ["Class"] ]
# store the variables we want to predicting on
target = "Class"
# for column in data.columns:
# if data[column].dtype == type(object):
# le = LabelEncoder()
# data[column] = le.fit_transform(data[column])
# X = data[column]
# X = data[column]
# Y = data[target]
X = data.drop(target, 1)
Y = data[target]
# Print the shapes of X & Y
print(X.shape)
print(Y.shape)
# define a random state
state = 1
# # define the outlier detection method
# clf = IsolationForest(
# max_samples=20, random_state=state)
classifiers = {
"Isolation Forest": IsolationForest(max_samples=len(X),
contamination=outlier_fraction,
random_state=state),
"Local Outlier Factor": LocalOutlierFactor(
n_neighbors = 20,
contamination = outlier_fraction)
}
# fit the model
n_outliers = len(Fraud)
for i, (clf_name, clf) in enumerate(classifiers.items()):
# fit te data and tag outliers
if clf_name == "Local Outlier Factor":
y_pred = clf.fit_predict(X)
print("LOF executed")
scores_pred = clf.negative_outlier_factor_
# Export the classifier to a file
with open('model.pkl', 'wb') as model_file:
pickle.dump(clf, model_file)
else:
clf.fit(X)
scores_pred = clf.decision_function(X)
y_pred = clf.predict(X)
print("IF executed")
# Export the classifier to a file
with open('model.pkl', 'wb') as model_file:
pickle.dump(clf, model_file)
# Reshape the prediction values to 0 for valid and 1 for fraudulent
y_pred[y_pred == 1] = 0
y_pred[y_pred == -1] = 1
n_errors = (y_pred != Y).sum()
# run classification metrics
print('{}:{}'.format(clf_name, n_errors))
print(accuracy_score(Y, y_pred ))
print(classification_report(Y, y_pred ))
And here is the final output of this model:
Local Outlier Factor:115
0.9997806549072266
precision recall f1-score support
0 1.00 1.00 1.00 524231
1 0.00 0.00 0.00 57
avg / total 1.00 1.00 1.00 524288
After that, when I try to get a prediction by using a flask service it returns [1] for both(Normal and Fraudulent) cases.
What I did wrong in this model?
What can I be improved? Is there something wrong with my data?
Help me, please!
Thanks in advance!
