Investigating Credit Card Fraud Detection

Motivation:

In present days payment systems are changed into online transactions. There are many types of online payment method like e-cash, credit card, internet banking etc. Credit card is one of the best method for online payment. Credit card is a plastic card which is issued to user for payment. Credit cards are mostly used in everywhere recent days. Using of credit cards are increasing day by day. It allowed to buy goods or other services. Anyone need not to carry any cash by using credit card. That’s why, every banks provide credit card to their customer. Every banks set ATM booth in many place because cardholder can withdraw money from anywhere. In present days online shopping are increasing and their payment system is online.

Credit cards are increasing day by day and fraudulent activities are also increasing day by day. Cardholders don’t know how fraudsters collect their information. Criminal activities are done taking these sensible information. Fraudster always try new technology and tactics to commit illegal things. Fraudster withdraw money from card without the knowledge of cardholders or banks. Sometimes fraudster steal a card, if user doesn’t know about this than it can cause financial loss for user. User doesn’t even know when, how fraudster use his/her card to withdraw money. This fraudulent activities should be stopped.

Literature Review:

Fraud detection problem are essential for bank to reduce their losses. There are various method for detecting fraudulent of credit card. Genetic Algorithms are used to detect or predict fraudulent transactions. It is used number of usage of card, location of usage card, balance available of a card, daily spending amount from card for detecting fraudulent activities. Genetic algorithms reduce the customer and bank from great losses. The speed of Genetic Algorithms is good, accuracy is medium and cost is low. One of the best classifier algorithms is k-nearest neighbor algorithms (KNN) incoming transaction is calculated for nearest point to new incoming transaction and the incoming transaction indicates fraud. K-nearest neighbor is very accurate and efficient. The speed of k-nearest neighbor is good, accuracy is medium and cost is expensive. Fraudulent of credit card is detected by decision tree which follows Luhn’s Algorithms. Luhn’s algorithms is used to validate card number and address matching rules check billing address and shipping address match or not. The effectiveness and correctness of this algorithms is secure. The speed of Decision Tree is very fast, accuracy is medium and cost is high. Neural Network (ANN) is latest technique to detect and predict fraudulent of credit card. Back Propagation Network (BPP) is popular learning algorithms to train neural networks. Back propagation network used for choosing parameter like weight, network type, number of layer, number of node etc. Genetic Algorithm is used with combining Neural Network. By using Genetic Algorithm and Neural Network (GANN) is together, the chances of success is very high [5]. The speed of Neural Network is very fast, accuracy is medium and cost is high. Behavior based classification approach using Support Vector Machine is very effective. Support vector machine is used in pattern recognition and classification. If any discrepancies happen in behaviors transaction pattern then it is predicted as fraud. Support vector machine use kernel to gains flexibility in the form of threshold for separating the data’s. Support vector machine gives good result, higher accuracy of detection. The speed of Support vector machine is very low, accuracy is medium and cost is high. Hidden Markov Model (HMM) is a stochastic process which is trained with the normal behavior of a cardholder. If an incoming credit card transaction is not accepted by HMM, it is considered fraud. Hidden Markov model suggest a method for finding spending profile of cardholder which is detect whether incoming transaction is fraudulent or not. The accuracy of Hidden Markov model is close to 80%. The speed of Hidden Markov model is very high, accuracy is medium and cost is very high. Another credit card fraud detection technique is outlier based on distance sum which proposes detection procedure and experiment with transaction data set. Outlier mining credit card fraud better than clustering detection. This algorithms apply in bank credit card fraud detection and the fraud transaction is predicted soon. This method prevents banks from losses and reduces rick fraudulent [8]. Bagging Ensemble Classifier based on decision tree algorithm which is an effective technique for credit card fraud detection and it uses real life dataset on credit card transaction. Bagging Ensemble Classifier based on decision tree algorithm works very well and takes less time to execute. Less time is one of the important parameter for real time application [9]. Machine Learning algorithms are detecting credit card fraudsters. Hybrid method using AdaBoost and majority voting methods uses publicly available credit card data set. Hybrid method using AdaBoost and majority voting methods analyze real world credit card data set from financial institution. Noise are added to the data sample and the majority voting method is stable in presence of noise. This method achieves good accuracy for detecting credit card fraud detection. A Big Data Analytical framework process large volume of data and extract data from different sources. The extracted data are used to build strong analytical model and to improve the analytical model three different analytical techniques are implemented and these analytical models are based on credit card dataset. Classification and prediction are very essential for credit card fraudulent [12].

Behavior pattern of a customer detect the credit card frauds. Behavior pattern of a customer are in train data set. Small numbers of data gives good result. The problem is if massive amount of data is given, Support Markov Model does not give good result. To avoid this problem extraction method is used which is used for data reduction. By choosing proper values for the parameters C and miu avoids over fitting and gives a perfect accuracy [6].

The above discussion of the research is how fraudulent of credit card is detected. There are various ways to detect or predict fraudulent of credit card describing in the research. Some method has problem some has not. Every method is effective for detecting fraudulent of credit card. But there are many questions that arise with these research: *Is every method is effective for our country?