fraud detection using anneecs.csuohio.edu/~sschung/cis601/frauddetectionnn...with simulated...

26
FRAUD DETECTION USING NEURAL NETWORKS SAIYAM KOHLI (2669077)

Upload: others

Post on 16-Aug-2020

3 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: FRAUD DETECTION USING ANNeecs.csuohio.edu/~sschung/CIS601/FraudDetectionNN...with simulated annealing algorithm. As the result shows that the training time is high but the fraud detection

FRAUD DETECTION USING

NEURAL NETWORKS

SAIYAM

KOHLI

(2669077)

Page 2: FRAUD DETECTION USING ANNeecs.csuohio.edu/~sschung/CIS601/FraudDetectionNN...with simulated annealing algorithm. As the result shows that the training time is high but the fraud detection

AGENDA

Types of Credit Card Fraud?

What is Artificial neural network?

SIMULATED ANNEALING

TRAINING OF ANN

RESULTS

CONCLUSION

Page 3: FRAUD DETECTION USING ANNeecs.csuohio.edu/~sschung/CIS601/FraudDetectionNN...with simulated annealing algorithm. As the result shows that the training time is high but the fraud detection

What is Credit Card Fraud?

Page 4: FRAUD DETECTION USING ANNeecs.csuohio.edu/~sschung/CIS601/FraudDetectionNN...with simulated annealing algorithm. As the result shows that the training time is high but the fraud detection

TECHNIQUE FOR FRAUD DETECTION

1)SUPERVISED: Use training data to build the models which we have attribute

of class label.

2)UNSPERVISED: Training data does not have class label.

Page 5: FRAUD DETECTION USING ANNeecs.csuohio.edu/~sschung/CIS601/FraudDetectionNN...with simulated annealing algorithm. As the result shows that the training time is high but the fraud detection

WHAT IS NEURAL NETWORK?

Page 6: FRAUD DETECTION USING ANNeecs.csuohio.edu/~sschung/CIS601/FraudDetectionNN...with simulated annealing algorithm. As the result shows that the training time is high but the fraud detection

WHAT IS NEURAL NETWORK?

Similar functionality like human brain.

Consist of artificial neurons which can be viewed as set of nodes in a network.

Application in business failure prediction, stock price prediction , credit card

fraud detection and many more area.

Page 7: FRAUD DETECTION USING ANNeecs.csuohio.edu/~sschung/CIS601/FraudDetectionNN...with simulated annealing algorithm. As the result shows that the training time is high but the fraud detection

FEED FORWARD NEURAL NETWORK

Page 8: FRAUD DETECTION USING ANNeecs.csuohio.edu/~sschung/CIS601/FraudDetectionNN...with simulated annealing algorithm. As the result shows that the training time is high but the fraud detection

SIMPLE FEED-FORWARD NETWORK

Page 9: FRAUD DETECTION USING ANNeecs.csuohio.edu/~sschung/CIS601/FraudDetectionNN...with simulated annealing algorithm. As the result shows that the training time is high but the fraud detection

PERCEPTRON FUNCTION IN NEURAL

NETWORK

INPUT FUNCTIONS: Collects all input and perform summation and transfer to

activation function.

ACTIVATION FUNCTIONS : Perform some operation on the result after

summation and transfer to the next level.

Page 10: FRAUD DETECTION USING ANNeecs.csuohio.edu/~sschung/CIS601/FraudDetectionNN...with simulated annealing algorithm. As the result shows that the training time is high but the fraud detection

VISUALIZATION OF FUNCTIONS

Page 11: FRAUD DETECTION USING ANNeecs.csuohio.edu/~sschung/CIS601/FraudDetectionNN...with simulated annealing algorithm. As the result shows that the training time is high but the fraud detection

EVALUATION OF SUMMATION

FUNCTIONS

Page 12: FRAUD DETECTION USING ANNeecs.csuohio.edu/~sschung/CIS601/FraudDetectionNN...with simulated annealing algorithm. As the result shows that the training time is high but the fraud detection

ACTIVATION FUNCTIONS

Result of summation function is passed to activation function ,which will scale

the value of S in a proper range.

Two types of Activation Function:

1)Sigmond Function: Works on threshold ,if the value of S crosses the threshold

then the node is pass as an output.

2)Hyberbolic Tangent Function:Next version of sigmoid function

Page 13: FRAUD DETECTION USING ANNeecs.csuohio.edu/~sschung/CIS601/FraudDetectionNN...with simulated annealing algorithm. As the result shows that the training time is high but the fraud detection

SIGMOND FUNCTION

REPRESENTATION:

Page 14: FRAUD DETECTION USING ANNeecs.csuohio.edu/~sschung/CIS601/FraudDetectionNN...with simulated annealing algorithm. As the result shows that the training time is high but the fraud detection

Hyperbolic Tangent activation function

REPRESENTATION:

Page 15: FRAUD DETECTION USING ANNeecs.csuohio.edu/~sschung/CIS601/FraudDetectionNN...with simulated annealing algorithm. As the result shows that the training time is high but the fraud detection

ANNEALING

Annealing is a thermal process for obtaining low energy states of a solid in a heat bath.

The process contains three steps:

1. Heat the system at high temperature T and generate a random solution.

2. As the algorithm progress, T decreases at each iteration and each iteration

forms a nearby model.

3. Then cool the system slowly until the minimum value of T is reached and

generate a model at each iteration, which takes the system towards global

minima.

Page 16: FRAUD DETECTION USING ANNeecs.csuohio.edu/~sschung/CIS601/FraudDetectionNN...with simulated annealing algorithm. As the result shows that the training time is high but the fraud detection

PROCEDURE OF SIMULATED ANNEALING

The main definitions which is needed for this algorithm are:

a method is to generate initial solution, by generating worst solution at the

beginning helps to avoid converging to local minimum

Perturbation Function to find a next solution with whom the current solution is

compared.

an Objective Function is to be defined to evaluate and rate the current solution

on the basis of performance,

an Acceptance Function, which is used to check whether the current solution is

good or not in comparison with the current one, a very basic one is

exp((currentSol-nextSol)/currentTemp).

the last one is stopping criteria, there are many stopping criteria’s, in this paper

we have used an threshold value of objective function as an stopping criteria.

Page 17: FRAUD DETECTION USING ANNeecs.csuohio.edu/~sschung/CIS601/FraudDetectionNN...with simulated annealing algorithm. As the result shows that the training time is high but the fraud detection

TRAINING OF ANN

Page 18: FRAUD DETECTION USING ANNeecs.csuohio.edu/~sschung/CIS601/FraudDetectionNN...with simulated annealing algorithm. As the result shows that the training time is high but the fraud detection

CALCULATIONS

Page 19: FRAUD DETECTION USING ANNeecs.csuohio.edu/~sschung/CIS601/FraudDetectionNN...with simulated annealing algorithm. As the result shows that the training time is high but the fraud detection

ANNEALING ALGORITHM

Page 20: FRAUD DETECTION USING ANNeecs.csuohio.edu/~sschung/CIS601/FraudDetectionNN...with simulated annealing algorithm. As the result shows that the training time is high but the fraud detection

RANDOMIZATION OF WEIGHTS

Page 21: FRAUD DETECTION USING ANNeecs.csuohio.edu/~sschung/CIS601/FraudDetectionNN...with simulated annealing algorithm. As the result shows that the training time is high but the fraud detection

RESULTS SET

Page 22: FRAUD DETECTION USING ANNeecs.csuohio.edu/~sschung/CIS601/FraudDetectionNN...with simulated annealing algorithm. As the result shows that the training time is high but the fraud detection

PARAMETERS OF ARTIFICIAL NEURAL

NETWORKS

Page 23: FRAUD DETECTION USING ANNeecs.csuohio.edu/~sschung/CIS601/FraudDetectionNN...with simulated annealing algorithm. As the result shows that the training time is high but the fraud detection

PARAMETERS OF SIMULATED ANNEALING

ALGORITHM

Page 24: FRAUD DETECTION USING ANNeecs.csuohio.edu/~sschung/CIS601/FraudDetectionNN...with simulated annealing algorithm. As the result shows that the training time is high but the fraud detection

RESULT OF TRAINED NEURAL NETWORKS

Page 25: FRAUD DETECTION USING ANNeecs.csuohio.edu/~sschung/CIS601/FraudDetectionNN...with simulated annealing algorithm. As the result shows that the training time is high but the fraud detection

CONCLUSION

In this paper we showed that better result is achieved with ANN when trained

with simulated annealing algorithm. As the result shows that the training time

is high but the fraud detection in real time is considerably low and the

probability of predicting the fraud case correctly in online transaction is high,

which is a main measure to evaluate any ANN.

The main problem in credit card fraud detection is the availability of real

world data for the experiment.

This approach can also be used in other applications which require

classification task [20] e.g. software failure prediction, etc

Page 26: FRAUD DETECTION USING ANNeecs.csuohio.edu/~sschung/CIS601/FraudDetectionNN...with simulated annealing algorithm. As the result shows that the training time is high but the fraud detection

THANK YOU