Credit Card Fraud Detection Using Fuzzy Rule Based Classifier
- Pages: 7
- Word count: 1530
- Category: Credit Card
A limited time offer! Get a custom sample essay written according to your requirements urgent 3h delivery guaranteed
Order NowFraud detection is an emerging topic of notable importance. Data mining strategies have been applied most considerably to the detection of insurance fraud, monetary fraud and financial fraud. This project will mainly focus on detecting fraudulent credit card transactions. Fraud detection in telecommunication systems, particularly the case of extraordinary imposed fraud, providing an anomaly detection technique supported by way of a signature schema, fraud deals with cases regarding criminal purposes that typically are different to identify, have additionally attracted a a tremendous deal of attention in latest years.
The use of credit cards has dramatically increased because of a fast advancement inside the electronic commerce technology. Credit card will become the most popular mode of payment for each on line as properly as ordinary purchase, in instances of fraud related to it are also growing day through day. In this research sequence of operations in credit card transaction processing using a Fuzzy rule based classifier and accuracy is improved in the detection of frauds compared to other algorithms. A Naïve Bayes is initially trained with the everyday behaviour of a card holder. If an incoming credit card transaction is not accepted by the trained version with sufficiently excessive probability, it’s considered to be fraudulent. At the same time, it ensures that true transactions aren’t rejected. Supervised learning requires prior type to anomalies. In this research fuzzy rule primarily based category set of rules used for modelling real world credit card information statistics and detecting the anomaly usage of credit card information’s. Whenever anomaly credit card usage detected the system will capture the anomaly user face and freeze the anomaly user system. Django framework is used for web app creation.
Keywords: Fuzzy Rule based Classifier, Naïve Bayes classifier, Data mining & Django Framework.
I. Introduction
Online Shopping has become the new age of shopping. It is expected that in the year 2021, over 2.14 billion people are supposed to buy goods and services online. As the number of online transactions increases the crime rate also increases. Credit card fraud events take place frequently and results in huge financial loss. Fraudulent transactions can occur in various ways and can be put into different categories such as Credit Card Fraud, Refund Fraud, Merchant Fraud and Identity theft etc. As the damage is higher in these fraudulent cases we have proposed a solution to overcome the fraudulent crimes and to identify the victim performing these crimes
Following are the modules used in this project to detect and overcome the credit card frauds. Authentication Module is a plug-in that collects user information such as a user ID and password, and compares the information against entries in a database. Load Dataset converts unstructured data into structured data format by writing data into database. Data Pre-processing, the data can have many irrelevant and missing parts. To handle this part, data cleaning is done. It involves handling of missing data, noisy data etc. Feature Extraction, in this module, a Fuzzy rule-based feature engineering that extracts both individual behaviour and frequency based behaviour in order identify the fraudulent behaviors. Fraudulent Detection, in this module fraudulent credit card behaviour will be detected by using fuzzy rule based classifier.
The project focuses on the implementation of Fuzzy rule based classifier. The fuzzy algorithm which uses supervised learning mechanism produces high accuracy when compared with the previous algorithms used in the base paper. Django framework is used to create a web application with the help of python libs. SQL LITE is used as a server and all these will be used in the Feature Extraction module to identify the fraudulent behaviours.
II. Literature Review
SARA MAKKI el.at (2018) In this paper they have explored these solutions by using machine learning algorithm for fraud detection. They have identified their weaknesses and summarized the results that they have obtained using a credit card fraud labeled dataset. According to this, the imbalanced classification approaches are ineffective especially when data are highly imbalanced. The existing approaches result in a large number of false alarms, which are costly to the financial institutions. This may lead to inaccurate detection as well as increasing the occurrence of fraud cases.
Algorithm: Imbalanced Classification
Changjun Jiang el.at (2018) ,they have proposed a window-sliding strategy to aggregate the transactions in each group. Next, they extract a collection of specific behavioral patterns for each cardholder based on the aggregated transactions and the cardholder’s history of transactions. Then they train a set of classifiers for each group on the base of all behavioral patterns. Finally, they use the classifier set to detect fraud online and if a new transaction is fraudulent, a feedback mechanism is taken in the detection process in order to solve the problem of concept drift.
Algorithm/Method: Sliding Window
Zahra Kazemi el.at (2017) Recently, several methods of data mining seek to find the best way for this prediction. One of the most popular methods that have been very useful in the past decade is employing deep learning. There are several methods of deep learning that can extract most applicable and abstract features form raw data. As auto encoders are so simple and they can be used regardless of class labels, they are so useful for our purpose. There are several variants of auto encoders that each one has pros and cons. regarding the AI of our problem we have used two of them in this research and compare the results with the state of the arts. Another advantage of employing deep auto encoders in this realm is according to big data maintenance.
III. Data Source
Kaggle is the world’s largest data science community with powerful tools and resources to help you achieve data science goals. They offer a large amount of datasets to be used in various projects to prove accuracy and lot more. In this project we have used a dataset with data of 284806 user’s behaviour profiles online. These behaviour profiles will help the algorithm to predict the fraudulent transaction accurately.
IV. Methodology
A. Fuzzy Classifier
A fuzzy classifier algorithm can be described by the way of a rule based in the following way. On the left-hand side of each rule, there may be a combination of values of linguistic variables that defines a specific class. The integer-valued variable, which identifies the identical elegance, is at the right –hand side. For any item described through values of its characteristics, its assignment to a category is decided by means of correspondence between these values and the left-hand side of the rules. The output of the fuzzy classification system depends on numeric class identifiers form a nominal, ordinal or maybe cardinal scale. Incase of a nominal scale, there is no use of using any arithmetic operations on magnificence identifiers. The result is processed into records whether the item is categorised or not. If the numeric elegance identifiers form a uniform cardinal evaluating scale, its miles meaningful to perform calculations whose result may be the place of an object in among classes. In each those cases, a verbal description of the result is desirable.
B. Django Framework
Django is a free and open source web application framework written in Python. A framework is nothing more than collection of modules that make improvement easier. They are grouped together and it permits the user to create applications or websites from an existing source ,in place of from scratch. This is how websites even simple ones designed by means of a single person can still include advanced functionality like authentication support, control and admin panels, contact forms & comment boxes.
In other words, if you creating a website from scratch you would need to develop these components. Framework components are built in ,so just needed to configure them properly to match the site you want.
In this experiment, the django framework is used to create the sample shopping site which is used by the attacker for performing a fraudulent transaction. In the coding part of this django framework, the application Pushbullet is linked to the backend of the website. The same application is also installed on the mobile device of the cardholder. Both of them are connected with the API (Application Programming Interface) key. So this application will capture the face of the attacker and it will send it to the cardholder.
C. Experimental Result
When a fraudulent transaction takes place on the dummy shopping website created for this experiment purpose, the fuzzy algorithm will predict the fraudulent transaction by the given dataset obtained from the website Kaggle. The fuzzy classifier algorithm which is used in this experiment provides higher accuracy when compared to the other algorithms used in the base paper. The accuracy it provides is 0.998, whereas the accuracy of other algorithms like Random Forest, Imbalanced Classifier and Sliding window collectively are 0.984. Not only it will predict but it will freeze the transaction, log out from the system and it will capture the face of the attacker and sends it to the cardholder.
D. Conclusion & Future Work
In this experimental research, we have identified and stop the attacker from making a fraudulent transaction on a fake shopping site. This project has proven to have higher accuracy. In this project a dummy site is only used, but this experiment can be used for all the online shopping sites for identifying the fraudulent transaction as a future updation.