Detect fake profiles in online social networks using Random

Forest
In [54]: import sys
import csv
import datetime
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
from datetime import datetime
import sexmachine.detector as gender
from sklearn.preprocessing import Imputer
from sklearn import cross_validation
from sklearn import metrics
from sklearn import preprocessing
from sklearn.metrics import roc_curve, auc
from sklearn.ensemble import RandomForestClassifier
from sklearn.cross_validation import StratifiedKFold, train_test_split
from sklearn.grid_search import GridSearchCV
from sklearn.metrics import accuracy_score
from sklearn.learning_curve import learning_curve
from sklearn.metrics import classification_report
from sklearn.metrics import confusion_matrix
%matplotlib inline

function for reading dataset from csv files

In [55]: def read_datasets():

""" Reads users profile from csv files """
genuine_users = pd.read_csv("data/users.csv")
fake_users = pd.read_csv("data/fusers.csv")
# print genuine_users.columns
# print genuine_users.describe()
#print fake_users.describe()
x=pd.concat([genuine_users,fake_users])
y=len(fake_users)*[0] + len(genuine_users)*[1]
return x,y

function for predicting sex using name of person

 In [56]: def predict_sex(name):
sex_predictor = gender.Detector(unknown_value=u"unknown",case_sensitiv
e=False)
first_name= name.str.split(' ').str.get(0)
sex= first_name.apply(sex_predictor.get_gender)
sex_dict={'female': -2, 'mostly_female': -1,'unknown':0,'mostly_mal
e':1, 'male': 2}
sex_code = sex.map(sex_dict).astype(int)
return sex_code

function for feature engineering

In [57]: def extract_features(x):

lang_list = list(enumerate(np.unique(x['lang'])))
lang_dict = { name : i for i, name in lang_list }
x.loc[:,'lang_code'] = x['lang'].map( lambda x: lang_dict[x]).astype(i
nt)
x.loc[:,'sex_code']=predict_sex(x['name'])
feature_columns_to_use = ['statuses_count','followers_count','friend
s_count','favourites_count','listed_count','sex_code','lang_code']
x=x.loc[:,feature_columns_to_use]
return x

function for ploting learning curve

 In [60]: def plot_learning_curve(estimator, title, X, y, ylim=None, cv=None,
n_jobs=1, train_sizes=np.linspace(.1, 1.0, 5)):

plt.figure()
plt.title(title)
if ylim is not None:
plt.ylim(*ylim)
plt.xlabel("Training examples")
plt.ylabel("Score")
train_sizes, train_scores, test_scores = learning_curve(
estimator, X, y, cv=cv, n_jobs=n_jobs, train_sizes=train_sizes)
train_scores_mean = np.mean(train_scores, axis=1)
train_scores_std = np.std(train_scores, axis=1)
test_scores_mean = np.mean(test_scores, axis=1)
test_scores_std = np.std(test_scores, axis=1)
plt.grid()

plt.fill_between(train_sizes, train_scores_mean - train_scores_std,

train_scores_mean + train_scores_std, alpha=0.1,
color="r")
plt.fill_between(train_sizes, test_scores_mean - test_scores_std,
test_scores_mean + test_scores_std, alpha=0.1, colo
r="g")
plt.plot(train_sizes, train_scores_mean, 'o-', color="r",
label="Training score")
plt.plot(train_sizes, test_scores_mean, 'o-', color="g",
label="Cross-validation score")

plt.legend(loc="best")
return plt

function for plotting confusion matrix

In [61]: def plot_confusion_matrix(cm, title='Confusion matrix', cmap=plt.cm.Blue

s):
target_names=['Fake','Genuine']
plt.imshow(cm, interpolation='nearest', cmap=cmap)
plt.title(title)
plt.colorbar()
tick_marks = np.arange(len(target_names))
plt.xticks(tick_marks, target_names, rotation=45)
plt.yticks(tick_marks, target_names)
plt.tight_layout()
plt.ylabel('True label')
plt.xlabel('Predicted label')

function for plotting ROC curve

 In [62]: def plot_roc_curve(y_test, y_pred):
false_positive_rate, true_positive_rate, thresholds = roc_curve(y_tes
t, y_pred)

print "False Positive rate: ",false_positive_rate

print "True Positive rate: ",true_positive_rate

roc_auc = auc(false_positive_rate, true_positive_rate)

plt.title('Receiver Operating Characteristic')

plt.plot(false_positive_rate, true_positive_rate, 'b',
label='AUC = %0.2f'% roc_auc)
plt.legend(loc='lower right')
plt.plot([0,1],[0,1],'r--')
plt.xlim([-0.1,1.2])
plt.ylim([-0.1,1.2])
plt.ylabel('True Positive Rate')
plt.xlabel('False Positive Rate')
plt.show()

Function for training data using Random Forest

In [63]: def train(X_train,y_train,X_test):

""" Trains and predicts dataset with a Random Forest classifier """

clf=RandomForestClassifier(n_estimators=40,oob_score=True)
clf.fit(X_train,y_train)
print("The best classifier is: ",clf)
# Estimate score
scores = cross_validation.cross_val_score(clf, X_train,y_train, cv=5)
print scores
print('Estimated score: %0.5f (+/- %0.5f)' % (scores.mean(), scores.st
d() / 2))
title = 'Learning Curves (Random Forest)'
plot_learning_curve(clf, title, X_train, y_train, cv=5)
plt.show()
# Predict
y_pred = clf.predict(X_test)
return y_test,y_pred
In [64]: print "reading datasets.....\n"
x,y=read_datasets()
x.describe()

reading datasets.....

Out[64]: id statuses_count followers_count friends_count favourites_count

count 2.818000e+03 2818.000000 2818.000000 2818.000000 2818.000000

mean 5.374889e+08 1672.198368 371.105039 395.363023 234.541164

std 2.977005e+08 4884.669157 8022.631339 465.694322 1445.847248

min 3.610511e+06 0.000000 0.000000 0.000000 0.000000

25% 3.620867e+08 35.000000 17.000000 168.000000 0.000000

50% 6.162253e+08 77.000000 26.000000 306.000000 0.000000

75% 6.177673e+08 1087.750000 111.000000 519.000000 37.000000

max 1.391998e+09 79876.000000 408372.000000 12773.000000 44349.000000

In [65]: print "extracting featues.....\n"

x=extract_features(x)
print x.columns
print x.describe()

extracting featues.....

Index([u'statuses_count', u'followers_count', u'friends_count',

u'favourites_count', u'listed_count', u'sex_code', u'lang_code'],
dtype='object')
statuses_count followers_count friends_count favourites_count \
count 2818.000000 2818.000000 2818.000000 2818.000000
mean 1672.198368 371.105039 395.363023 234.541164
std 4884.669157 8022.631339 465.694322 1445.847248
min 0.000000 0.000000 0.000000 0.000000
25% 35.000000 17.000000 168.000000 0.000000
50% 77.000000 26.000000 306.000000 0.000000
75% 1087.750000 111.000000 519.000000 37.000000
max 79876.000000 408372.000000 12773.000000 44349.000000

listed_count sex_code lang_code

count 2818.000000 2818.000000 2818.000000
mean 2.818666 -0.180270 2.851313
std 23.480430 1.679125 1.992950
min 0.000000 -2.000000 0.000000
25% 0.000000 -2.000000 1.000000
50% 0.000000 0.000000 1.000000
75% 1.000000 2.000000 5.000000
max 744.000000 2.000000 7.000000
In [66]: print "spliting datasets in train and test dataset...\n"
X_train,X_test,y_train,y_test = train_test_split(x, y, test_size=0.20, ran
dom_state=44)

spliting datasets in train and test dataset...

In [67]: print "training datasets.......\n"

y_test,y_pred = train(X_train,y_train,X_test)

training datasets.......

('The best classifier is: ', RandomForestClassifier(bootstrap=True, clas

s_weight=None, criterion='gini',
max_depth=None, max_features='auto', max_leaf_nodes=None,
min_samples_leaf=1, min_samples_split=2,
min_weight_fraction_leaf=0.0, n_estimators=40, n_jobs=1,
oob_score=True, random_state=None, verbose=0, warm_start=Fals
e))
[ 0.93791574 0.93791574 0.94678492 0.9578714 0.93777778]
Estimated score: 0.94365 (+/- 0.00395)

In [68]: print 'Classification Accuracy on Test dataset: ' ,accuracy_score(y_test,

y_pred)

Classification Accuracy on Test dataset: 0.941489361702

In [70]: cm=confusion_matrix(y_test, y_pred)
print('Confusion matrix, without normalization')
print(cm)
plot_confusion_matrix(cm)

Confusion matrix, without normalization

[[265 3]
[ 30 266]]

In [71]: cm_normalized = cm.astype('float') / cm.sum(axis=1)[:, np.newaxis]

print('Normalized confusion matrix')
print(cm_normalized)
plot_confusion_matrix(cm_normalized, title='Normalized confusion matrix')

Normalized confusion matrix

[[ 0.98880597 0.01119403]
[ 0.10135135 0.89864865]]
In [72]: print(classification_report(y_test, y_pred, target_names=['Fake','Genuin
e']))

precision recall f1-score support

Fake 0.90 0.99 0.94 268

Genuine 0.99 0.90 0.94 296

avg / total 0.95 0.94 0.94 564

In [73]: plot_roc_curve(y_test, y_pred)

False Positive rate: [ 0. 0.01119403 1. ]

True Positive rate: [ 0. 0.89864865 1. ]