Web Personalization Using Feedforward Backpropagation

4033 words (16 pages) Essay

26th Mar 2018 Computer Science Reference this

Disclaimer: This work has been submitted by a university student. This is not an example of the work produced by our Essay Writing Service. You can view samples of our professional work here.

Any opinions, findings, conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of UKEssays.com.

WEB PERSONALIZATION USING FEEDFORWARD BACKPROPAGATION NEURAL NETWORK

Chapter 4: Methodology

Chapter 4 shows the methodology of the present work. Section 4.1 present the methodology, section 4.2 includes flow chart of the present work. Section 4.3 present the proposed algorithm.

1.1 METHODOLOGY

  • Start
  • Configure search engine
  • Training on data according to user’s context.
  • Testing on data
  • ANN optimization for search optimization.
  • Stop

1.1

Training: – Data is trained using Feedforward Backpropagation Neural network. Before testing and searching the data is trained. Data training is required for optimal results.

Testing: – Testing of data is performed using Feedforward Backpropagation neural network and Using SVM (Support Vector Machine).

User Query: – User can enter the Query for find the information. Some Web sites name are suggested to the users according the query of user.

1.2 FLOWCHART

The simple flowchart of the designed algorithm is depicted in the figure 4.2

Figure 4.2: flowchart of the designed algorithm

1.3 ALGORITHM DESIGN

The Algorithm for the present work is discussed in this section. The various steps used in algorithm are explained. The pseudo Code for the algorithm is also discussed in this section.

Table4.1 Proposed Algorithm

Proposed Algorithm

STEP 1:

(GENERATION OF WT(TRAINING AND TESTING)

Begin

Initialize CAPS chars(‘A’,’Z’)

Initialize small chars(‘a’ to ‘z’)

For each char in training/testing file

Find(pos~char)

T(pos.char)

STEP 2:

(SUPPORT VECTOR MACHINE)

Begin

test_set=gettest data;

For each category in cat Test_Set=generator(Test_Set)

Training _set(category, :)=wt(category, :); Test_Set=unique(Test_Set);

For j=1:wt.count Kernal.function=linear;

Group(j)=j; result=MultiSVM(Ts,Gruop,Test_Set)

End

Ts=find(unique(training.set));

STEP 3:

Begin

Initialize Training Set As SVM

Group=GroupSVM;

Neural.Initialization=newt(TrainSet,Group);

Train.hiddenNeurons=10;

(hidden neurons may vary according to size)

Train.epocs=50;

Result_train=train(net,trainset,group);

Result_class=simulate(result_train,test_data);

STEP 4:

(ERROR CALCULATION)

Error=sqrt((sum(result_class)-sum(original_class))^2/length(result)

   
  1. Pseudo code of proposed algorithm:-The Pseudo code for proposed algorithm is shown below.

Table 4.2 proposed algorithm in pseudo code

Algorithm_Design

Start

globaltesting_datauser_found

extracting the words from the paragraph

result=generator(current_word)

load architecture

Select A Training File To Upload

collecting data from the excel sheet

for setting up the target;

fori=1:rows

for k=1:cols

Target(j)=p;

j=j+1;

end

p=p+1;

end

initiating the neural network

loadsvm_group

loadtraining_data;

fortempind=1:itrind

tst=test(tempind,:);

C=Cb;

T=Tb;

u=unique(C);

N=length(u);

c4=[];

c3=[];

j=1;

k=1;

if(N>2)

itr=1;

classes=0;

cond=max(C)-min(C);

while((classes~=1)&&(itr<=length(u))&& size(C,2)>1 &&cond>0)

if you increase the data you will have to adjust the groups also

Database Updated

Draw Plots

End

Chapter-5

RESULT AND PERFORMANCE ANALYSIS

In this chapter results of the present is explained. The figures of result, comparison, comparison tables and graphs of the present work are shown in this chapter.

1.1 TOOLS USED

To implement my work I used Matlab. Matlab Stands for MATrix LABoratory. MATLAB has a modern programming language environment: it has refined data structures, contains built-in editing and debugging tools, and supports object-oriented programming.

Table 5.1: Tools Used

Computer

Core i 3 or higher

RAM

32 MB

Platform

Windows xp/ 7/8

Other hardware

Keyboard, mouse

Software

Matlab 2010a

MATLAB

The name MATLAB stands for MATrix LABoratory. MATLAB was written originally to provide easy access to matrix software developed by the LINPACK (linear system package) and EISPACK (Eigen system package) projects MATLAB is a high-performance language for technical computing. It integrates computation, visualization, and programming environment. Furthermore, MATLAB is a modern programming language environment: it has refined data structures, contains built-in editing and debugging tools, and supports object-oriented programming. These factors make MATLAB an outstanding tool for education and research. MATLAB has many advantages compared to conventional computer languages (e.g., C, FORTRAN) for solving technical problems. MATLAB is an interactive system whose basic data element is an array that does not require dimensioning. The software package has been commercially available since 1984 and is now considered as a standard tool at most universities and industries worldwide. It has powerful built-in routines that enable a very wide variety of computations. It also has easy to use graphics commands that make the visualization of results immediately available. Specification applications are collected in packages referred to as toolbox. There are toolboxes for signal processing, symbolic computation, control theory, simulation, and optimization.

Get Help With Your Essay

If you need assistance with writing your essay, our professional essay writing service is here to help!

Find out more

After logging into your account, you can enter MATLAB by double-clicking on the MATLAB shortcut icon (MATLAB 7.0.4) on your Windows desktop. When you start MATLAB, a special window called the MATLAB desktop appears. The desktop is a window that contains other windows. The major tools within or accessible from the desktop are:

  • The Command Window
  • The Command History
  • Workspace
  • The Current directory
  • Help browser
  • Start button

5.1.1 MATLAB CHARACTERISTICS

  • Developed first and foremost by Cleve Molar in the 1970’s
  • Derived from FORTRAN subroutines LINPACK and EISPACK, linear and Eigen value systems.
  • Developed principally as an interactive system to access LINPACK and EISPACK.
  • Gained its esteem through word of mouth, because it was not authoritatively dispersed.
  • Rewritten in C in the 1980’s with more functionality, which include plotting routines.
  • The Math Works Inc. was produced (1984) to marketplace and go on with expansion Of MATLAB.

5.1.2 ADVANTAGES OF MATLAB

  • MATLAB may behave as a calculator or as a programming language
  • MATLAB combine adequately calculation and graphic plotting.
  • MATLAB is moderately easy to learn
  • MATLAB is interpreted (not compiled), errors are easy to fix.
  • MATLAB is optimized to be relatively fast when performing matrix operations
  • MATLAB does have some object-oriented elements

5.1.3 RESULTS

In this section Screen Shots of the present work are shown. Firstly, Data Set is uploaded after that Neural Network and SVM are used for training and testing of the data. User can create their account and if user has already account then he can sign in for the Personalization. Three parameters are taken for the comparison between the SVM (support vector machine) and Neural Network. Accuracy, Precision and Recall are the three parameters used for the comparison. Neural Network gives the best results.

Figure 5.1: Proposed Flowchart

Fig. 5.1 shows the main working window of the personalization. The above figure has all the training and testing window components in w

+hich the personalized data can be trained through the Neural Network and Support Vector Machine.

Training Model for SVM as well as Neural Network.

Inputs: examples, a set of examples, each with input x = x1; x2; : : : ; xn and output y

Inputs: network, a perceptron with weights Wj ; j = 0; : : : ; n and activation function g

Repeat for each e in examples do

inPnj = 0Wj xj [e] Err y[e] – g(in)

WjWj + _ _ Err _ g0(in) _ xj [e]

End

Until all examples correctly predicted or stopping criterion is reached Return network

Figure 5.2: represents the architecture of the Neural Network

Neural network contains of input and hidden layers. Each and every layer has weight and bandwidth of the data. Hidden Layer contains epochs that means iteration. The maximum iteration provided over here is 50 but it is not necessary that the neural will run till 50. It would cross check the validations and would provide the results required. The results can also be checked by the following graphs.

Figure 5.3: Representing detailed neural architecture

The above figure represents the architecture over which the neural has been tested and trained. There is one validation denoted by the pink line and has been achieved on the 4th Iteration.

Figure 5.4: Personalizing Option

The above figure provides the option to personalize the system according to the choice of the user. Here the user can banned those website link which he or she does not want to see in the future.

Figure 5.5: Login window

Figure 5.5 shows the login window. If the user is new or not registered then he can use sign up option for registration. After filling details, user is registered. Useris alreadyregisteredhe can log in using theUserID and Password.

Figure 5.6: Results after testing data

The above figure represents results after testing the data. User can test data after fill the data in the box.

Figure 5.7. Different parameters

The above figure shows the different parameter after click on result neural button. Accuracy, Precision, and recall parameters can be calculated. The same parameter can be calculated by SVM also.

5.2Comparison Tables and Graphical Representation

The experiment was conducted for computing Accuracy, Precision and Recall. The experiment has been performed to compare the performance of both Neural Network and SVM (Support Vector Machine). The Accuracy, Precision and Recall for both approaches was different. Given tables and graphs proves the performance of the algorithms.

Table 5.2: Accuracy Comparison

ACCURACY COMPARISON

CATEGORY

ACCURACY NEURAL

ACCURACY SVM

POLITICS

97.9348

90.3484

ENTERTAINMENT

99.1023

96.4148

SPORTS

97.9348

89.3067

HOUSE

99.5281

96.76

EDUCATION

95.696

91.742

GOOD

93.7454

98.558

BEAUTIFUL

96.641

94.1688

HAPPY

90.7599

99.0201

FOOD

96.7325

88.1569

FRUITS

90.3681

95.5261

VEGETABLES

85.4341

92.3119

COLORS

95.8125

89.3325

CLOTHES

97.1374

99.1023

TOOLS

96.4148

94.4251

SHAPES

99.5209

89.3036

     

Figure 5.8: Graph of accuracy comparison

Table 5.3: Precision Comparison

PRECISION COMPARISON

CATEGORY

PRECISION NEURAL

PRECISION SVM

POLITICS

1.2828

1.3548

ENTERTAINMENT

1.1053

1.124

SPORTS

1.2828

1.2674

HOUSE

1.891

1.376

EDUCATION

1.0934

1.1997

GOOD

1.164

1.1472

BEAUTIFUL

1.588

1.3946

HAPPY

1.0874

1.2957

FOOD

1.1206

1.0738

FRUITS

1.1985

1.2635

VEGETABLES

1.1878

1.2608

COLORS

1.0489

1.0409

CLOTHES

1.4406

1.1053

TOOLS

1.124

1.1562

SHAPES

1.4564

1.2803

     

Figure 5.9: graph of Precision Comparison

Table5.4: Recall Comparison

CATEGORY

RECALL NEURAL

RECALL SVM

POLITICS

0.792

0.67773

ENTERTAINMENT

0.89813

0.84409

SPORTS

0.792

0.68945

HOUSE

0.67537

0.89335

EDUCATION

0.85077

0.74268

GOOD

0.95213

0.86386

BEAUTIFUL

0.81413

0.84414

HAPPY

0.77884

0.80129

FOOD

0.85091

0.74772

FRUITS

0.72527

0.9208

VEGETABLES

0.66707

0.87509

COLORS

0.88184

0.78242

CLOTHES

0.73272

0.89813

TOOLS

0.84409

0.96881

SHAPES

0.68564

0.89048

Figure 5.10: Graph of Recall Comparison

Chapter 6: CONCLUSION AND FUTURE SCOPE

Chapter 6 includes conclusion and future scope of the present work. Future scope means that what enhancement can be done in the future. Section 6.1 covers the Conclusion and Section 6.2 covers the Future scope.

6.1 CONCLUSION

Web personalization is an answer for data over-burden issue on World Wide Web .The web personalization assemble the accuracy of web hunt apparatus, streamlines the looking process and reduce the time customer needs to spend for looking for. Today for both Web-based affiliations and for the end customers the web personalization has transformed into a key gadget.

Find out how UKEssays.com can help you!

Our academic experts are ready and waiting to assist with any writing project you may have. From simple essay plans, through to full dissertations, you can guarantee we have a service perfectly matched to your needs.

View our services

Web utilization mining is the methodology of recognizing delegate patterns and scanning examples depicting the movement in the site, by investigating the clients’ conduct. Site directors can then use this information to redesign or change the site according to the side interests and behaviour of its visitors, or upgrade the execution of their systems. Also, the supervisors of e-trade destinations can procure profitable business brainpower, making buyer profiles and accomplishing business sector division. There exists number of techniques yet none has been accomplished great amount.

This postulation introduced a methodology taking into account neural system for web personalization of web substance. Firstly, in the pre-processing stage the information must be gathered from the better places it is put away (customer side, server side, and intermediary servers). In the wake of recognizing the customers, the snap surges of each customer must be part into sessions. The last venture of the entire web utilization mining methodology is to dissect the examples found amid the example disclosure step. Web Usage Mining attempt to comprehend the examples identified in before step. The most well-known systems is information visualization applying channels High dimensional information stream contains a huge colossal measure of information. Such huge sum information contains a vast information with high measurements with information many-sided quality. A valid example remote sensor framework data, web logs, Google look for, et cetera. Standard strategies are not suitable over high dimensional data as they obliged high figuring expense for taking care of data that is the reason this technique has been realized with some change highlights.

6.2 FUTURE SCOPE

Future misleads examine the half breed utilization structure positioning that can be connected to a bound together web/navigational diagram which extends out of the breaking points of a solitary site. Such approach would empower a “worldwide” significance positioning over the web, improving both web query items and the suggestion process.

Now, if the user wants to revisit URL P3, she would not be able to do that using just the BackButton navigation Stack. If she resorts to the history list to get some help, she will be disappointed to see that its list based textual representation gives no idea about the structure of the navigation pattern. Moreover, even for a modestly sized navigation session, the history list gets cluttered to an extent so that renders it ineffective in searching for a specific page. The bookmark facility is of little help in this case, as the user cannot bookmark each and every page due to overhead associated with the very process of bookmarking. Moreover, even selected bookmarking is of no help as, in most cases, the user does not know at the time of visiting a web page whether it is important enough to be bookmarked.

One thing that has long been acknowledged by the research community is the use of graphical overview diagrams in assisting user navigation through complex information spaces.

The visualization scheme employed should be efficient enough to give a graphical representation of user session history in real time. Computationally and graphically intensive application may cause undue delays in the visualization generation process, especially when the session history grows large. Most of the past work done for WWW subspace visualization is plagued by these delays therefore is inefficient for the ordinary use.

The solution must be designed keeping in mind that it has to replace WWW browser stack based navigation structure and its history list. Therefore it must provide all those facilities that were provided by these browser components. Users who are familiar with the facilities provided by the browser may find it very difficult to adjust to a new scheme that does not provide these facilities.

The visualization scheme should be designed more on an aesthetic rather than a scientific basis. Humans tend to get confused when presented with a large amount of data jumbled up in front of them. It is, therefore, highly recommended that session history data be divided into small and easily manageable groups, neatly knitted together through an elegant link structure.

1

Cite This Work

To export a reference to this article please select a referencing stye below:

Reference Copied to Clipboard.
Reference Copied to Clipboard.
Reference Copied to Clipboard.
Reference Copied to Clipboard.
Reference Copied to Clipboard.
Reference Copied to Clipboard.
Reference Copied to Clipboard.

Related Services

View all

DMCA / Removal Request

If you are the original writer of this essay and no longer wish to have your work published on the UKDiss.com website then please: