Fraud Detection Using Python | A Guide For Developers By Raghib Khan

1
Fraud Detection Using Python

• A Guide For Developers By Raghib Khan

2
Table Of Contents

• Why Using Python?
• Machine Learning Algorithms Used In Python
• Process
• Importing Necessary Libraries
• Importing Dataset
• Data Processing Understanding
• Train Test Split
• Model Building

3

Why Using Python?
4

• Cybersecurity is becoming increasingly important
  in our daily lives as we transition to a digital
  world. Finding unusual behavior is the
  fundamental problem when discussing digital life
  security.
• Many people often prefer credit cards when they
  transact online or buy something. Credit card
  credit limits occasionally enable us to make
  purchases even when we don't have the money on
  hand.
• On the other hand, online criminals abuse these
  features. To solve this issue, we require a
  system that can halt a transaction if it detects
  irregularities. Hence comes machine learning to
  the rescue!
• There are various ML techniques available now
  that can classify unusual transactions and Python
  makes use of maximum number of techniques to find
  fraudulent transactions.
• It only requires historical data and an algorithm
  that can match our data more effectively. Here, a
  system monitors all transactions' patterns is
  required, and if any patterns are unusual, the
  transaction should be stopped.

5
Machine Learning Algorithms In Python
6
Raghib Khan shares that his developers at Phonato
Studios uses these five algorithms of machine
learning in Python

• Naive Bayes
• Decision tree
• Linear regression
• Logistic regression
• Support Vector Machines (SVM)

7



Process
8

• Raghib Khan, who is against fraud and fraudulent
  activities, shares the process of utilizing
  Python in understanding which transactions are
  genuine and which are not!
• Lets check the process involved!

9
Importing Necessary Libraries Dataset ?
10

• Co-founded by Raghib Khan and Faisal Abdi, RNF
  Technologies and its team believe in making quick
  modifications. Importing all the required
  libraries to one location is a good idea.
• Consider that you have credit card data the
  features in the dataset are the PCA's transformed
  versions, so you won't need to go through the
  feature selection process again.
• To determine the best features for your model in
  any other case, it is advised to employ RFE,
  RFECV, SelectKBest, and VIF scores.
• The dataset import process is relatively
  straightforward, and you can use python's Pandas
  module to import it.
• Run the command shown below to import your data.
• datapd.read_csv("c
  reditcard.csv")

11
Data Processing Understanding ?
12

• The dataset is skewed in favor of one attribute,
  which is the one thing you will immediately
  notice about this data.
• It appears to be reasonably accurate for this
  type of data.
• Since many institutions currently use various
  security measures, it is more difficult for
  hackers to carry out such actions.
• However, the likelihood of such activities can
  occasionally rise when the system is vulnerable.
• The fact that the bulk of transactions in the
  datasets are normal and just a tiny fraction of
  transactions are fraudulent allows you to see
  this.

13
Train Test Split ?
14

• Raghib Khan, leader of fraud detection team at
  Phonato Studios, suggests that before splitting
  the train and test phases, you must specify the
  dependent and independent variables.
• Both the dependent and independent variables go
  by the names X and Y.
• X data.drop("Class," axis 1).values
• y data"Class".values
• Let's divide our train and test data now.
• Y train, Y test, X train, and X test are equal to
  train test split(X, y, test size 0.25, random
  state 1).
• You currently have two distinct data sets the
  train data, which you will use to train the
  model, and the test data, which will be derived
  from the unseen data.

15
Model Building ?
16

• You can choose from various machine learning
  models depending on what is most convenient for
  you.
• The model definition is significantly more
  straightforward. A single line of code can define
  your model and fit the model to the data.
• These models can also be tuned by choosing
  various optimum parameters.
• But it does not need to be complicated if the
  accuracy is greater even with less parameter
  adjustment.
• Congratulations! You have successfully utilized
  python language to detect credit card fraud with
  the help of the web development wizard Raghib
  Khan.

17
Special Thanks To

• Raghib Khan,
• Co-Founder,
• RNF Technologies