# Idea About Fraud Detection Using Different Algorithms Computer Science Essay

Published:

This essay has been submitted by a student. This is not an example of the work written by our professional essay writers.

The first algorithm that the author has used is mining the symbolic data. This algorithm is based on the idea that misused transaction are seen as a kind of rule. The advantage of this is, Combining a number of misuse rules which leads us to shorten the rules and decrease the dependency. The other rules is called mining analog data. This rules is based on dealing with analog data. Here the problem of fraud analysis is based on separating two kinds of classes of events. This algorithm leads us to high fraud detection and high confidence. In last the author used another technique by combining rule based association algorithm and network information in decision network. When this system is used in parallel then this shows error in large amount along with low confidence. Beside that this system is used in sequential as well. Now the advantage of this system is that the analog data first passes through analog check and then moves forward for sequential check , which leads us to high data correctness and high confidence as well.

Author name

Paper name

Published date

Paper description

Algorithm used

No of algorithm

No of inputs

Run time efficiency

R.Bruce, T.langsdorf, M.hepp, frankfur A. M

Neural data mining for credit card detection

## Ã-

IEEE

Mining the symbolic data , mining the analog data, combining symbolic and analog data

3

## Ã-

## Ã-

Time complexity

Memory

Future idea

Synthetic data

Real time

Tools for experiment

Experimental environment

Tree structure is used

Language for encoding

## Ã-

## Ã-

Statistic based credit card fraud detection

## âˆš

## Ã-

A Neural network algorithm

## Ã-

## âˆš

## Ã-

Pruned network

Advantage

Related model/algorithm

Based on

Type of data

Algorithm type

Data presentation

## Ã-

High correctness and high confidence

Traditional vs advanced

Decisions

## Ã-

## Ã-

Tables and graphs

1) R.Bruce, T.langsdorf, M.hepp, frankfur A. M. Neural data mining for credit card detection

[2] This paper is based on effective data mining using neural network. The paper shows an approach to find out symbolic classification set of laws by means of neural network. Whereas neural network is first qualified to attain the obligatory precision pace. Furthermore the unneeded link of network are detached by mean of network pruning process. The activation principles of the unseen element in the system are evaluated. Which leads us to the classification rules of particular analysis. Research is conducted on this planned approach using a distinct set of data mining difficulty. And the outcome demonstrate that high quality of rules can be exposed from the known data sets.

Author name

Paper name

Published date

Paper description

Algorithm used

No of algorithm

No of inputs

Run time efficiency

Hongjun Lu, Rudysetiono huan liu

Effective data mining using neural network

1996

IEEE

Finding symbolic classification using neural network

3

38

## Ã-

Time complexity

Memory

Future idea

Synthetic data

Real time

Tools for experiment

Experimental environment

Tree structure is used

Language for encoding

## Ã-

Smaller in size

To generate rules similar to decision tree

## Ã-

## âˆš

Classification using neural network

## Ã-

## âˆš

## Ã-

Pruned network

Advantage

Related model/algorithm

Based on

Type of data

Algorithm type

Data presentation

## Ã-

To reduce training time of neural network

Extracting rules from trained neural network

Rules

## Ã-

Advanced

## Ã-

2) Hongjun Lu, Rudysetiono huan liu. 1996 Effective data mining using neural network

[3]This paper is based on adoptive neural network model for financial study. Data mining plays central role in finding buried predictive information from bulky catalog. Artificial neural network which is commonly used by data mining technique, is an algorithm used for this purpose. This proposed approach has been tested with function approximation and stock market moment study. And from these experimental data we can wind up that the future approach is much better to existing standard ANN for the use of data mining. And we can see that our new ANN with NAF can boost trained speed, shrink network size and simulation error. And also provide us more promising outcomes.

Author name

Paper name

Published date

Paper description

Algorithm used

No of algorithm

No of inputs

Run time efficiency

Dr Shuxiang xu, Prof Ming Zang

An adoptive neural network model for financial analysis

2005

Journal

Artificial neural network

2

## Ã-

## Ã-

Time complexity

Memory

Future idea

Synthetic data

Real time

Tools for experiment

Experimental environment

Tree structure is used

Language for encoding

## Ã-

## Ã-

To explore ANN with NAF for financial analysis

## âˆš

## âˆš

## Ã-

## Ã-

## âˆš

## Ã-

Pruned network

Advantage

Related model/algorithm

Application type

Based on

Type of data

Algorithm type

Data presentation

## Ã-

Reduce network size and simulation error

Neuron adoptive activation function

NAF and ANN

Rules

## Ã-

Advanced

Graphs

3) Dr Shuxiang xu, Prof Ming Zang. 2005 An adoptive neural network model for financial analysis

[4]this paper describes the implementation of artificial neural network in field of solid ducted rocket test. This paper describes the brief implementation of artificial neural network model. Further more it tells that ANN is combined with RBF(redial basis function) to recover the abnormal data. The ANN model is based on 3 layered architecture. Which takes data on input node then transform it to middle layer called hidden node then it moves toward the output node. When data moves toward the hidden node then it applies data mining techniques on large data and find the correlation between different data. This algorithm detects and recover the irregular parameters rapidly and efficiently.

Author name

Paper name

Published date

Paper description

Algorithm used

No of algorithm

No of inputs

Run time efficiency

Qiang liu, futting bao, bingge xia, xi'an P.R

Data analysis and SDR test based on ANN model

2012

## Ã-

Radial basis function

1

## Ã-

## Ã-

Time complexity

Memory

Future idea

Synthetic data

Real time

Tools for experiment

Experimental environment

Tree structure is used

Language for encoding

## Ã-

## Ã-

Still under process of development

## Ã-

## âˆš

Artificial neural network

SDR test

## Ã-

## Ã-

Pruned network

Advantage

Related model/algorithm

Application type

Based on

Type of data

Algorithm type

Data presentation

## Ã-

Detect and recover the abnormal parameter effectively

k-mean clustering algorithm

ANN layered based

Rules

Test data

## Ã-

## Ã-

4) Qiang liu, futting bao, bingge xia, xi'an P.R 2012 Data analysis and SDR test based on ANN model

[5] This paper tells us about the implementation of data mining approach in urban water system. In data mining approach further they implement neural network approach. Further more in detail they introduced self-organizing maps approach in neural network. The job of which is to collect DNA based molecular techniques and to analyze environmental samples. in microbiology to group different samples. Comparison of many T-RFLP(terminal restriction fragment length polymorphism) profiles to discovercollective and singlecomponents of microbiology community. T-Align software is used for grouping these things. The main benefit of this approach is the capability to present the data in a visual way that offers effortless visualization and understanding of multi-dimensional and complex data sets.

Author name

Paper name

Published date

Paper description

Algorithm used

No of algorithm

No of inputs

Run time efficiency

Stephen R. mounce,henriette S.jensen,Catherine A.biggs, joby B. boxall

T-RFLP profiles from urban water system sampling using SOM maps

2012

## Ã-

Artificial neural network using SOM map

1

## Ã-

## Ã-

Time complexity

Memory

Future idea

Synthetic data

Real time

Tools for experiment

Experimental environment

Tree structure is used

Language for encoding

## Ã-

## Ã-

To learn the growth of this approach for idea to distributed structure

## âˆš

## Ã-

T-align software

Urban water system

## Ã-

## Ã-

Pruned network

Advantage

Related model/algorithm

Application type

Based on

Type of data

Algorithm type

Data presentation

## Ã-

The skill to show the data in visual way, explanation of complex data sets

Traditional statistical methods

ANN layered based

Rules

Synthetic data

Supervised/unsupervised

Unseen to visual

5) Stephen R. mounce,henriette S.jensen,Catherine A.biggs, joby B. boxall 2012 T-RFLP profiles from urban water system sampling using SOM maps

[6] In this paper the author describes the unsupervised visual data mining using SOM and a data driven color mapping. The author uses two different algorithm for finding the solution of this problem. The 1st method that the author used is SOM(self-organizing map). This algorithm yield two dimensional and irregular representation of the input records. Blinchard's approach utilizes this data as input and associates this input data to a pixel in a figure. The algorithm is a well-organized way to visualize the typical and large data. Finally he make use of these mention algorithm collectively.Which leads us to obtain a entirely unverified visual data mining instrument. Where the color mapping is data driven. The testing outcome of this approach offers visualization that permit the taking out of cluster. The unverified automation of the coloring allows us to nuance the attachment of a class.

Author name

Paper name

Published date

Paper description

Algorithm used

No of algorithm

No of inputs

Run time efficiency

Cyril De Runz, Eric desjardin, Michel Herbin,crestic

Unsupervised data mining using SOM & data driven color mapping

2012

Journal

Kohonen map and blanchard approach

2

## Ã-

## Ã-

Time complexity

Memory

Future idea

Synthetic data

Real time

Tools for experiment

Experimental environment

Tree structure is used

Language for encoding

## Ã-

## Ã-

selecting unclear agronomical size and cooperate within GIS.

## Ã-

## âˆš

k-mean and MATlab

## Ã-

## Ã-

## Ã-

Pruned network

Advantage

Related model/algorithm

Application type

Based on

Type of data

Algorithm type

Data presentation

## Ã-

## Ã-

k-means, MATlab toolbox

Combination of SOM and color mapping

Decision/rule based

Real data unknown

Unsupervised

Visual

6) Cyril De Runz, Eric desjardin, Michel Herbin,crestic 2012 Unsupervised data mining using SOM & data driven color mapping

[7] The theme of the author is to make different clusters using data mining technique. He make use of self-organizing map(SOM) to accomplish his goal. The aim of SOM is to map multi-dimensional input into two dimensional form. This tactic is used for clustering and classification purpose. The rules are extracted from trained SOM's. which can then figure the prepositional IFâ€¦ THEN type system. These effortless set of laws can be easily broken by expert or decision support system and are simply interpretable to an expert. The law signify the trained SOM in cases where clustering has willingly taken place. The important feature of this projected scheme is the underlying exactness of clustering procedure achieved by SOM. In case where the SOM not succeed to readily cluster the facts, the ensuing system will replicate this original inexactness.

Author name

Paper name

Published date

Paper description

Algorithm used

No of algorithm

No of inputs

Run time efficiency

James Malone, Kenneth McGarry, Safan Wermter,Chris Buwerman

Data mining using rule extraction from kohhnen SOM

2005

Journal

Self-organizing maps

1

## Ã-

## Ã-

Time complexity

Memory

Future idea

Synthetic data

Real time

Tools for experiment

Experimental environment

Tree structure is used

Language for encoding

## Ã-

Smaller in size

## Ã-

## âˆš

## âˆš

## Ã-

## Ã-

## Ã-

## Ã-

Pruned network

Advantage

Related model/algorithm

Application type

Based on

Type of data

Algorithm type

Data presentation

## Ã-

vital accuracy of clustering route execute by SOM

Extracting rules

Combination of SOM and color mapping

Rule based

Real data

Unsupervised

Clustering of different data

7) James Malone, Kenneth McGarry, Safan Wermter,Chris Buwerman 2005 Data mining using rule extraction from kohhnen SOM

[8] In this paper the author describes a novel approach for Expert system application. He make use of an algorithm called MTS(Mahalanobis-Taguchi system)-ANN(artificial neural network) in expert system. He implements this algorithm in dynamic environment. The experimental outcomes of this algorithm prove that this algorithm is vastlyvalid in pattern recognition and is computationally efficient in addition to the ANN algorithm is a straightforward and resourcefulsystem for assembling a dynamic structure. From this it can be accomplished that MTS-ANN algorithm can be effectively useful to dynamic environment for data-mining troubles.

Author name

Paper name

Published date

Paper description

Algorithm used

No of algorithm

No of inputs

Run time efficiency

Ching-lien huang, tsung-shin hsu, chih-ming liu

The MTS-ANN algorithm for data mining in dynamic environments

2009

Journal

MTS-ANN

1

Large data sets

## Ã-

Time complexity

Memory

Future idea

Synthetic data

Real time

Tools for experiment

Experimental environment

Tree structure is used

Language for encoding

## Ã-

Smaller in size

MTS-ANN algorithm can be used in dynamic environment perfectly

## âˆš

## âˆš

Statistical data, charts

Dynamic environment

## Ã-

## Ã-

Pruned network

Advantage

Related model/algorithm

Application type

Based on

Type of data

Algorithm type

Data presentation

## Ã-

Pattern recognition, model construction and high confidence

Linear correlation discovering(LCD)

Combination of SOM and color mapping

Rule/statistical based

Real data

## Ã-

Charts and tables

8) Ching-lien huang, tsung-shin hsu, chih-ming liu 2009 The MTS-ANN algorithm for data mining in dynamic environments

[9] The given paper shows machine learning and data mining application for the prediction of drifts in technology skilled turnover rates of the employees. Then he used an algorithm which is the combination of two different algorithm i.e SOM(self organizing map) and BPN(back propagation neural network). This algorithm combines the advantages of SOM and BPN which applied on the expose properties associated to turnover trends cluster. With the help of this algorithm we come to know that this algorithm is the best algorithm for finding out the turnover of employees and also showing the factors which involve in increasing the rate of the employees turnover.

Author name

Paper name

Published date

Paper description

Algorithm used

No of algorithm

No of inputs

Run time efficiency

Chin-yaun fan, pei-shu Fan, te-ye chan, shu-hao shang

Using hybrid data mining and machine learning clustering analysis to predict the turnover rate for technology professional

2012

Journal

clustering analysis(SOM) plus Back propagation neural network

3

421 valid questioners

## Ã-

Time complexity

Memory

Future idea

Synthetic data

Real time

Tools for experiment

Experimental environment

Tree structure is used

Language for encoding

## Ã-

## Ã-

## Ã-

## âˆš

Using data mining tools

Using data mining methodology

## Ã-

## Ã-

Pruned network

Related model/algorithm

Based on

Type of data

Algorithm type

Data presentation

## Ã-

SOM and neural network clustering method

Rules

Real time data

Questioners

Charts and graphs

9) Chin-yaun fan, pei-shu Fan, te-ye chan, shu-hao shang 2012 Using hybrid data mining and machine learning clustering analysis to predict the turnover rate for technology professional

[10] In this proposed paper the author introduces a new neural approach known as ensemble recursive rule extraction. This approach is basically mining of rules from the ensemble neural network. In this approach we come to know that the proposed approach produces higher recognition accuracy as compared to the individual neural network. Where the mined rules are more comprehensible. The proposed approach gives more rules than the previous approaches. For neural data analysis this proposed approach promises a new approach. So it is clear that in future this approach will be used to escalate the opportunities to use data mining for the purpose of high data recognition.

Author name

Paper name

Published date

Paper description

Algorithm used

No of algorithm

No of inputs

Run time efficiency

Atsushi Hara, Yoichi Hayashi

New neural data analysis approach using ensemble neural network rule extraction

2012

ICAN

Ensemble Recursive Rules Extraction (E-Re-Rx)

1

## Ã-

## Ã-

Time complexity

Memory

Future idea

Synthetic data

Real time

Tools for experiment

Experimental environment

Tree structure is used

Language for encoding

## Ã-

Large amount on memory

Using data mining with high recognition accuracy

## Ã-

## âˆš

Using neural network rules

Getting of rules from ensemble neural network

## âˆš

## Ã-

Pruned network

Related model/algorithm

Based on

Type of data

Algorithm type

Data presentation

## Ã-

## Ã-

Rules

Learning data sets

## Ã-

## Ã-

10) Atsushi Hara, Yoichi Hayashi 2012 New neural data analysis approach using ensemble neural network rule extraction

[11] Dealing with the predictions of the winner of the college football team is a challenging and interesting task. The previous studies shows us that all the previous prediction were failed because they were dealing with the ranking and force of the team. Here the author has predicted a novel approach the author used three techniques(artificial neural network, support vector machine and decision trees) the purpose of using these technique is to create regression and classification kind of models so that to review different methodologies prediction ability.This method proved that this approach is better way to present the future predictions and can provide a lot of accurate results than the previous predictions.

Author name

Paper name

Published date

Paper description

Algorithm used

No of algorithm

No of inputs

Run time efficiency

Dursun Delen, Douglas Cogdell, Nihat Kasap

A comparative analysis of data mining methods in predicting NCAA bowl outcomes

2012

Journal

CRISP-DM

3

244 Bowl Games

## Ã-

Time complexity

Memory

Future idea

Synthetic data

Real time

Tools for experiment

Experimental environment

Tree structure is used

Language for encoding

## Ã-

## Ã-

Enrichment of variable set, identifying and including more input variable

## Ã-

## âˆš

Decision tree

Artificial neural network

Support vector machine

Predicting the outcomes of the college football games

## âˆš

## Ã-

Pruned network

Related model/algorithm

Based on

Type of data

Algorithm type

Data presentation

## Ã-

## Ã-

Rules and predictions

Real data

## Ã-

Chart and graphs

11) Dursun Delen, Douglas Cogdell, Nihat Kasap 2012 A comparative analysis of data mining methods in predicting NCAA bowl outcomes

[12] The author predict a new intelligent approach which will forecast the frequent growth of software which is base on theof the functional networks forecasting framework. There are lots of other methods which forecast the prediction of the software development. But all these failed because these have number of drawbacks like how to deal with the uncertainties. The planned approach has high tendency to deal with the grown environment of recent software progress. The result shows that this method is far better than then other approaches, its performance is sure and give us a smallest MAPE value.

Author name

Paper name

Published date

Paper description

Algorithm used

No of algorithm

No of inputs

Run time efficiency

Emad A. El-Sebakhy

Functional network as a novel data mining paradigm in forecasting software development effort

2011

Journal

Functional network intelligent system

1

## Ã-

## Ã-

Time complexity

Memory

Future idea

Synthetic data

Real time

Tools for experiment

Experimental environment

Tree structure is used

Language for encoding

## Ã-

## Ã-

To use different independent other than polynomial to use for data bases

## Ã-

## Ã-

Neural network and functional network

Forecasting to the software development efforts

## Ã-

Artificial neural network

Pruned network

Related model/algorithm

Based on

Type of data

Algorithm type

Data presentation

## Ã-

LOC, COCOMO, Function Point(FP)

predictions

## Ã-

Functional network forecasting frame work

Graph

12) Emad A. El-Sebakhy 2011 Functional network as a novel data mining paradigm in forecasting software development effort

[13] The motive of this paper is to deal with non-financial and financial ratios of the financial statements. For the association of performance of the financial distress prediction, he make use of the clustering and BPN modeling. So that to get an early alarm. From the results of the method he comes to know four major critical attributes. The 1st is that as more as we are using the factor analysis our result for clustering and BPN will be less accurate. 2nd is that that as soon as we get close to the actual financial distress, we will catch more precise outcome. 3rd is BPN has lower average rate of type ii errors as compare to the clustering model.In 4th the last stage the BPN provide a better and efficient prediction as that of the DM clustering approach.

Author name

Paper name

Published date

Paper description

Algorithm used

No of algorithm

No of inputs

Run time efficiency

Wei-Sen, Yin-kuan Du

Using neural networks and data mining techniques for the financial distress prediction model

2009

Journal

Artificial neural network and data mining techniques

1

Large amount of records

## Ã-

Time complexity

Memory

Future idea

Synthetic data

Real time

Tools for experiment

Experimental environment

Tree structure is used

Language for encoding

## Ã-

Large amount of memory

Using existing techniques to deal with more financial datasets

## âˆš

## âˆš

ANN and data mining

Financial distress environment

## Ã-

## Ã-

Pruned network

Related model/algorithm

Based on

Type of data

Algorithm type

Data presentation

## âˆš

BPN, clustering and classification

Rules and statistical calculation

Supervised/unsupervised

Supervised/unsupervised

Charts and graphs

13) Wei-Sen, Yin-kuan Du 2009 Using neural networks and data mining techniques for the financial distress prediction model

[14] The theme of the paper is to collect information about breast cancer disease. Which is known as a serious cancer disease through out the globe. The goal of this author is to divide woman in two categories i.e the woman who has broad confirmations of having breast cancer are grouped into malignant and having no breast disease are grouped into benign.So the author tried to propose such a hybrid breast cancerdiagnose system by joining together artificial neural network and MRAS. This method is then combine with the BPN. Now in this case this model has high classification accuracy. Form the result of the hybrid and the combination of hybrid and BPN it is clear that Hybrid and BPN provide lots of accuracy but the main advantage of hybrid system is that it can save lots of implementation time which results in making shorten the time for on time decisions.

Author name

Paper name

Published date

Paper description

Algorithm used

No of algorithm

No of inputs

Run time efficiency

Shienu-Ming Chou, Tian -Shyug Lee, Yuehjen E.Shao, I-Fei Chen

Mining the breast cancer pattern using artificial neural network and multivariate adaptive regression splines

2004

Data mining technique with multivariate adoptive regression splines(MARS)

1

## Ã-

## Ã-

Time complexity

Memory

Future idea

Synthetic data

Real time

Tools for experiment

Experimental environment

Tree structure is used

Language for encoding

## Ã-

Large amount

Collecting more important variable that increase classification accuracy

## âˆš

## âˆš

Artificial intelligence and data mining techniques

Woman health care

## Ã-

## Ã-

Pruned network

Related model/algorithm

Based on

Type of data

Algorithm type

Data presentation

## Ã-

BPN

Rules and decisions

## Ã-

## Ã-

Charts and graphs

14) Shienu-Ming Chou, Tian -Shyug Lee, Yuehjen E.Shao, I-Fei Chen 2004 Mining the breast cancer pattern using artificial neural network and multivariate adaptive regression splines

[15] In this research the author is curious in the field of transportation and wants to apply different data mining techniques to find out dissimilarities, similarities. He is cautious to know similarities and dissimilarities between two different school of thoughts. He furthermore made experiments on different technique using complex structure. And find out that along with the advantage of the complex modeling tools it has limitation as well. Which is a big hurdle in the way of finding similarities and dissimilarities. From this he simply concluded that instead of using complex modeling technique the goal of analyzing is much important. Because there are always assumption in all modeling approaches. So with the help of simpler model give us a good result just like complex one.

Author name

Paper name

Published date

Paper description

Algorithm used

No of algorithm

No of inputs

Run time efficiency

M.G Karlatis, E.I Vlahogianni

Statistical method vs neural networks in transportation research, differences, similarities and some insights

2011

Journal

Statistical and computational thoughts.

2

## Ã-

## Ã-

Time complexity

Memory

Future idea

Synthetic data

Real time

Tools for experiment

Experimental environment

Tree structure is used

Language for encoding

## Ã-

## Ã-

## Ã-

## âˆš

Statistical predictions and artificial intelligence

Transportation research

## Ã-

## Ã-

Pruned network

Related model/algorithm

Based on

Type of data

Algorithm type

Data presentation

## Ã-

## Ã-

Predictions and previous data

Supervised/unsupervised

## Ã-

charts

15) M.G Karlatis, E.I Vlahogianni 2011 Statistical method vs neural networks in transportation research, differences, similarities and some insights

[16] In this article the author focuses on one of major issue in neural network. Neural network have been used for regression and classification method in past. And the interpretation of their internal representation were very difficult. Now a day, it is clear that for the extraction of the understandable representation from trained neural network algorithm can be derived. The purpose of which is to use for data mining applications. The work mentioned in this paper delivers a generalized procedure, which can be used for the problems in bioinformatics. The results are really impressive but the problem is that of producing a large amount of data. Combination of algorithm with the neural network for the purpose of extracting of information from the trained neural network is the best solution, which produces high accurate data along with that produces knowledge discovery. The usage of these methods leads to acceptance and high confidence.

Author name

Paper name

Published date

Paper description

Algorithm used

No of algorithm

No of inputs

Run time efficiency

Antoney Browne, Brain D. Hudson, David C. Whitley, Martyn G. Ford, Philip Picton

Biological data mining using neural networks implementation and application of a flexible decision tree extraction algorithm to genomic problem domains

2003

article

Statistical and computational thoughts.

2

## Ã-

## Ã-

Time complexity

Memory

Future idea

Synthetic data

Real time

Tools for experiment

Experimental environment

Tree structure is used

Language for encoding

## Ã-

## Ã-

## Ã-

## âˆš

Statistical predictions and artificial intelligence

Transportation research

## âˆš

Matlab/Netlab

Pruned network

Related model/algorithm

Based on

Type of data

Algorithm type

Data presentation

## Ã-

Trepan Algorithm

Decision tree

## Ã-

Traditional vs new techniques

tables

16) Antoney Browne, Brain D. Hudson, David C. Whitley, Martyn G. Ford, Philip Picton 2003 Biological data mining using neural networks implementation and application of a flexible decision tree extraction algorithm to genomic problem domains

[17] In this article the author is dealing with the prediction of gully initiation. In past predicting gully initiation was prepared with the help of GIF scheme with knowledge base expert system, physical based system or statistical procedures. But while applying these procedures validity and reliability are big issues. For the identification and risk of gully initiation a procedure known as Data mining which is based on decision trees is applied. In this article the comparison of DM technique is shown with many other procedures like expert system and topographic threshold method (TT). The results show that DM technique provides more accurate data than that of the other methods. So it is obvious that for the study of the erroneous procedure and gully initiation a valuable technique is DM technique.

Author name

Paper name

Published date

Paper description

Algorithm used

No of algorithm

No of inputs

Run time efficiency

Tal Svoray, Evgenia Michailov, Avraham Cohen, Lior Rokah and Arnon Sturm

Predicting gully initiation:Comparing data mining techniques, analytical hierarchy processes and the topographic

2012

Article

Data mining procedure based on decision trees

1

10

## Ã-

Time complexity

Memory

Future idea

Synthetic data

Real time

Tools for experiment

Experimental environment

Tree structure is used

Language for encoding

## Ã-

Large amount

## âˆš

## âˆš

## Ã-

Transportation research

## âˆš

## Ã-

Pruned network

Related model/algorithm

Based on

Type of data

Algorithm type

Data presentation

## Ã-

Decision tree algorithm, ANN, TT,AHP

Rules and statistical analysis

Test data

Traditional vs DM techniques

Graph, charts and tables

17) Tal Svoray, Evgenia Michailov, Avraham Cohen, Lior Rokah and Arnon Sturm 2012 Predicting gully initiation:Comparing data mining techniques, analytical hierarchy processes and the topographic

[18] The aim of this paper is to examine the scalability of PNN (probabilistic neural network) through localization, a chain gradient tuning and parallelism. As PNN model is working in parallel so three well know approaches are studied here. Two fast approximation solutions are proposed by author in this paper. The main aim of this paper is to accelerate the PNN model with the help of 24 processors. And the result obtained reveals that PNN training along with subtractive clustering approaches and cross validation can amazingly escalate 24 times. The 2nd issue is how to eliminate the sigma parameter without major loss in PNN performance. Clustering within classes the most representative points are selected. Which results a localized PNN having tiny pattern Neuran Size and excellent performance, which is 10 times faster as compare to that of original version. For the most excellent PNN architecture, tuning can performed to using chain gradient to test it.

Author name

Paper name

Published date

Paper description

Algorithm used

No of algorithm

No of inputs

Run time efficiency

Yiannis Kkkinos and Konstantinos Margaritis

Parallelism, Localization and Chain Gradient Tuning Combination for Fast Scalable Probabilistic Neural Network in data mining

2012

Investigation of probabilistic neural network using localization and parallelization and a chain gradient tuning

3

## Ã-

## âˆš

Time complexity

Memory

Future idea

Synthetic data

Real time

Tools for experiment

Experimental environment

Tree structure is used

Language for encoding

O(n)

## Ã-

## âˆš

## Ã-

Artificial neural network and data mining

## Ã-

## Ã-

C using MPI library

Pruned network

Related model/algorithm

Based on

Type of data

Algorithm type

Data presentation

## Ã-

bayesian

Rules

## Ã-

## Ã-

Graphs and tables

18) Yiannis Kkkinos and Konstantinos Margaritis 2012 Parallelism, Localization and Chain Gradient Tuning Combination for Fast Scalable Probabilistic Neural Network in data mining

[19] In this whole paper the author is curious about getting high blood pressure data from a hospital data base. He is using back propagation algorithm in a multi-layered neural network. In building decision the results offered are very attractive. With the help of artificial intelligence approaches the hypothesis or activities of data is shown along with that formerly unknown data is exposed. This is all due to neural networks that the author is able to model irregular data and complex structure along with that various other issues.

Author name

Paper name

Published date

Paper description

Algorithm used

No of algorithm

No of inputs

Run time efficiency

Mbuyi Mukendi Kafunda Katatayi Pierre, Mbuyi Badibanga Sreve, Mbuyi Mukendi Didier

Extraction knowledge from high pressure data patients

2012

Journal

Multi-layer neural network

2

## Ã-

## Ã-

Time complexity

Memory

Future idea

Synthetic data

Real time

Tools for experiment

Experimental environment

Tree structure is used

Language for encoding

## Ã-

## Ã-

## Ã-

## âˆš

Neural network using supervised learning

Extraction of knowledge form data base with high blood pressure

## Ã-

## Ã-

Pruned network

Related model/algorithm

Based on

Type of data

Algorithm type

Data presentation

## Ã-

BPN

rules

Supervised

## Ã-

Tables

19) Mbuyi Mukendi Kafunda Katatayi Pierre, Mbuyi Badibanga Sreve, Mbuyi Mukendi Didier 2012 Extraction knowledge from high pressure data patients

[20] In this article the author focuses on one of major issue in neural network. Neural network have been used for regression and classification method in past. And the interpretation of their internal representation were very difficult. Now a day, it is clear that for the extraction of the understandable representation from trained neural network algorithm can be derived. The purpose of which is to use for data mining applications. The work mentioned in this paper delivers a generalized procedure, which can be used for the problems in bioinformatics. The results are really impressive but the problem is that of producing a large amount of data. Combination of algorithm with the neural network for the purpose of extracting of information from the trained neural network is the best solution, which produces high accurate data along with that produces knowledge discovery. The usage of these methods leads to acceptance and high confidence.

Author name

Paper name

Published date

Paper description

Algorithm used

No of algorithm

No of inputs

Run time efficiency

Antoney Browne, Brain D. Hudson, David C. Whitley, Martyn G. Ford, Philip Picton

Implementation and application of a flexible decision tree extraction algrothm to genomi problem domain

2004

Journal

Statistical and computational thoughts.

1

## Ã-

## Ã-

Time complexity

Memory

Future idea

Synthetic data

Real time

Tools for experiment

Experimental environment

Tree structure is used

Language for encoding

## Ã-

## Ã-

## Ã-

## Ã-

## âˆš

Statistical predictions and artificial intelligence

Genomic problem domain

## âˆš

Matlab/Netlab

Pruned network

Related model/algorithm

Based on

Type of data

Algorithm type

Data presentation

## Ã-

Trepan Algorithm

Decision tree

## Ã-

Traditional vs new approach

Tables

20) Antoney Browne, Brain D. Hudson, David C. Whitley, Martyn G. Ford, Philip Picton 2004 Implementation and application of a flexible decision tree extraction algrothm to genomi problem domain

[21] The author is dealing with a major issue of prediction for software quality. For resolving this hurdle he introduces an interpretable neural network model. This model consists upon a three layered feed forward neural network having sigmoid in its buried units. The output unit is having identity function and the model is trained accordingly. From trained neural network for the extraction of the rules he make use of the clustering genetic algorithm. For the detection of the fault prone software the rules extracted from the trained neural network are gathered. And then the rules are compared. The rules of trained neural network are compared with the rule of predicting results. The results show that trained neural network rules a bit accurate as compare to that of predicting results but predicting results are more understandable.

Author name

Paper name

Published date

Paper description

Algorithm used

No of algorithm

No of inputs

Run time efficiency

Qi wang. Bo yu, jie zhu

Extract Rules from Software Quality Prediction Model Based on

Neural Network

2004

IEEE

Clustering genetic and modeling methodology

2

## Ã-

## Ã-

Time complexity

Memory

Future idea

Synthetic data

Real time

Tools for experiment

Experimental environment

Tree structure is used

Language for encoding

## Ã-

## Ã-

## Ã-

## âˆš

Clustering and artificial intelligence

Software development efforts

## Ã-

C language

Pruned network

Related model/algorithm

Based on

Type of data

Algorithm type

Data presentation

## Ã-

Fitness function and cross over mutation

Rules

Training data

## Ã-

Tables

21) Qi wang. Bo yu, jie zhu 2004 Extract Rules from Software Quality Prediction Model Based on Neural Network

[22] In data mining classification is a key subject to be focused. But dealing with incomplete survey then classification is an innovative subject. Traditional neural network and other technique did not focus on the incomplete survey. So the author presents a novel approach known as extension neural network approach to deal with the incomplete survey. This proposed approach is dealing with the supervised data. After comparing the result of the planned approach with other approaches, it clearly shows that this approach has the benefit of high accuracy over other approaches.

Author name

Paper name

Published date

Paper description

Algorithm used

No of algorithm

No of inputs

Run time efficiency

Chao Lu, Xue- Wei Li, Hong-Bo Pan

Application of Extension Neural Network for Classification with Incomplete

Survey Data

2006

IEEE

Extension neural network based on model of clustering

1

## Ã-

## Ã-

Time complexity

Memory

Future idea

Synthetic data

Real time

Tools for experiment

Experimental environment

Tree structure is used

Language for encoding

## Ã-

Small amount

develop an

unsupervised learning algorithm of the proposed

extension neural network,

## Ã-

## âˆš

ANN and data mining technique

classification with incomplete survey

## Ã-

## Ã-

Pruned network

Related model/algorithm

Based on

Type of data

Algorithm type

Data presentation

## Ã-

Neural network and classification

Rules

supervised

New approach

Graphs

22) Chao Lu, Xue- Wei Li, Hong-Bo Pan 2006 Application of Extension Neural Network for Classification with Incomplete Survey Data

[23] In this paper the author is dealing with different application of the data mining and then selects the best method among them. Firstly he finds out the hidden pattern using different data mining approaches. Then according to their results he Kept based on Back propagation neural network, the result of which is providing improved security as compare to the rest of them.

Author name

Paper name

Published date

Paper description

Algorithm used

No of algorithm

No of inputs

Run time efficiency

L. Wang and T. Z. Sui

Application of Data Mining Technology Based on

Neural Network in the Engineering

2007

IEEE

Back Propagation neural network algorithm

1

## Ã-

## Ã-

Time complexity

Memory

Future idea

Synthetic data

Real time

Tools for experiment

Experimental environment

Tree structure is used

Language for encoding

## Ã-

## Ã-

## Ã-

## âˆš

Artificial neural network

Engineering

## Ã-

Matlab

Pruned network

Related model/algorithm

Based on

Type of data

Algorithm type

Data presentation

## Ã-

## Ã-

Rules and decisions

Supervise/unsupervised

## Ã-

Tables and graphs

23) L. Wang and T. Z. Sui 2007 Application of Data Mining Technology Based on Neural Network in the Engineering

[24] The author focuses on the diagnosis and fault forecast methods of stream turbine which is totally based on the genetic and neural network. For fault diagnosis in stream turbine and in data mining the genetic and neural network algorithm were introduces. By comparing the novel approach to the traditional approach it is quite clear that has better performance and simple to design. With the help of proposed algorithm sharing turbine fault diagnosis and rules extraction which is based on genetic and neural network algorithm are producing good results. The given system provide a good output with better confidence along with that it has strong ability of fault tolerant.

Author name

Paper name

Published date

Paper description

Algorithm used

No of algorithm

No of inputs

Run time efficiency

Qingling

A novel approach of diagnoses stream turbine based on neural network and genetic algorithm

2008

IEEE

Novel approach based on neural network and genetic algorithm

2

## Ã-

## Ã-

Time complexity

Memory

Future idea

Synthetic data

Real time

Tools for experiment

Experimental environment

Tree structure is used

Language for encoding

## Ã-

Large amount of memory

## Ã-

## âˆš

Artificial neural network with genetic algortithm

Stream turbine

## Ã-

## Ã-

Pruned network

Related model/algorithm

Based on

Type of data

Algorithm type

Data presentation

## âˆš

## Ã-

Rules

## Ã-

Traditional vs novel approach

Tables

24) Qingling 2008 A novel approach of diagnoses stream turbine based on neural network and genetic algorithm

[25] in this paper the author focuses on the ANN and finds out different terms which are responsible for making ANN system more intelligent. The author is looking forward and finding different ways for optimizing ANN based on classification methods. In this paper the author makes use of three standard datum for computing the accuracy. The experimental results show that the system gave high accuracy.

Author name

Paper name

Published date

Paper description

Algorithm used

No of algorithm

No of inputs

Run time efficiency

Guan Ping

Exploitation of Minimum Risk System based on Artificial Neural Network

2011

IEEE

Artificial neural network

1

## Ã-

## Ã-

Time complexity

Memory

Future idea

Synthetic data

Real time

Tools for experiment

Experimental environment

Tree structure is used

Language for encoding

## Ã-

Large amount of memory

## Ã-

## âˆš

## Ã-

Artificial neural network

To know about ANN intelligence

## Ã-

## Ã-

Pruned network

Related model/algorithm

Based on

Type of data

Algorithm type

Data presentation

## âˆš

## Ã-

Rules and decisions

Supervise/unsupervised

## Ã-

Tables

25) Guan Ping 2011 Exploitation of Minimum Risk System based on Artificial Neural Network