Methods

Advantages

Disadvantages

Speed of Detection

Accuracy

Cost

Comments

Neural Network

1) Able to learn from the past.

2) Extract rules and predict future activity based on the current situation.

3) Detect real time credit card fraud.

1) A number of parameter has to be set before any training can begin. There are no clear rules how to set these parameters.

Fast

Medium

Expensive

This method can be used when previous knowledge and past experience is known to the machine.

Genetic Algorithm

1) Works well with noisy data.

2) Easy to integrate with other system.

3) Usually combined into other techniques to increase the performance of these techniques and optimized their parameters.

1) The fraud is detected which is relevant to the customer’s behavior. A new classification problem which has inconsistent misclassification cost is introduced.

2) Requires extensive tool knowledge to set up and operate and difficult to understand.

Good

Medium

Inexpensive

This method can be used to detect or predict the fraud in a very short period of time and to minimize the number of false alerts.

Hidden Markov Model

1) Can detect the fraudulent activity at the time of the transaction.

1) Cannot detect fraud with a few transactions.

2) Produces high false alarm as well as high false positive.

3) Not scalable to large size data set.

Fast

Low

High expensive

It has been very effective for more complicated stochastic process.

Bayesian Network

1) Needs training of data to operate and require high processing speed.

2) More accurate and much faster than neural network.

1) Excessive training need.

2) BBNs are slower when applied to new instances.

Very fast

High

Expensive

Very effective for modeling situations where some information is already known and incoming data is unsure or partially unavailable.

Decision Tree

1) High flexibility/Good haleness/Explainable/ Easy to implement/Easy to display and to understand.

1) Requirements to check each condition one by one.

Fast

Medium

Expensive

This method has been effective for classification or regression.

Clustering Method

1) Clustering helps in grouping the data into similar clusters that helps in uncomplicated retrieval of data.

1) There are quite a few non-fraudulent activities which wrongly got detected as frauds. So to detect fraud accurately and efficiently it is necessary that the real data should be available.

High

Medium

Expensive

This method is formed to detect fraud in credit card transaction which are low, high, risky and high risky.

Self-Organizing Map (SOM)

---

---

Fast

Medium

Expensive

This method has been very effective when machine has no previous knowledge.