The E-business sector is rapidly evolving and the needs for web market places that anticipate the needs of the customers and the trust towards a product are equally more evident than ever. While people are enjoying the benefits from online trading, criminals are also taking advantages to conduct fraudulent activities against honest parties to obtain illegal profits. Therefore the requirement for predicting user needs and trust providence in order to improve the usability and user retention of a website can be addressed by personalizing and using a fraud product detection system.
Hence fraud-detection systems are commonly needed to be applied to detect and prevent such illegal or untrusted products. In this, we propose an online model framework which takes online feature selection, coefficient bounds from human knowledge and multiple instances learning into account simultaneously. By empirical experiments on a real-world we show that this model can potentially meet user needs, calculate the trust for a product and significantly reduce customer complaints.
1. PROACTIVE MODERATION AND A
PERSONALISED SYSTEM FOR FRAUD
PRODUCT DETECTION
Under the Esteemed Guidance of
MS.G. JYOTHI
(Assistant Professor)
By
K.SUNIL (10L35A1202)
P. RAMA LAKSHMI (09L31A1232)
PRABHA TETA (09L31A1234)
J.KARTHIK (09L31A1219)
2. ABSTRACT
The E-business sector is rapidly evolving and the needs for web market
places that anticipate the needs of the customers and the trust towards a
product are equally more evident than ever. While people are enjoying the
benefits from online trading, criminals are also taking advantages to conduct
fraudulent activities against honest parties to obtain illegal profits. Therefore
the requirement for predicting user needs and trust providence in order to
improve the usability and user retention of a website can be addressed by
personalizing and using a fraud product detection system.
3. Hence fraud-detection systems are commonly needed to be applied
to detect and prevent such illegal or untrusted products. In this, we
propose an online model framework which takes online feature selection,
coefficient bounds from human knowledge and multiple instances
learning into account simultaneously. By empirical experiments on a
real-world we show that this model can potentially meet user needs,
calculate the trust for a product and significantly reduce customer
complaints.
4. INTRODUCTION
Fraud detection and web personalization are the two key technologies
needed in various e-business applications to,
• Manage customer organization relationships
• Promote products
• Manage Web site content
• Provide knowledge to the user about the product.
The objective of this application is to “provide users with the
trustworthy products they want or need”.
5. Name : Proactive Moderation and A personalized System for Fraud
Product Detection
Purpose : To make user available time with trust worthy products
without Spending much of the time in knowing about
the product
Inputs : Ratings, Feedback
Outputs : Trusty worthy products are made available
Security : Usernames and password to each user
User Interface : Buttons and links on the screen allow the user to control the
system.
REQUIREMENT SPECIFICATION
The following are the functional and non functional Requirements
5
6. PROCEDURE
The phases of this process are:
Collection of data
The data to analyze is all about whether to trust the product or not so
the data will be
• Feedback from customer about the product
• Where the product has not meet the customer needs like
poor services/manufacturing
product mismatch
not delivered
Product damaged
7. Analysis of the collected data
The ways that are employed in order to analyze the collected data
include
Rule-based features:
Human experts with years of experience created many rules to
detect whether a user is fraud or not. It checks whether the product has
been or complained as untrusting or fraud.
The trust for particular product(X) can be calculated by
Trust(X)=100-Fraud(X)
Fraud(X)=No of complaints(X)/(No of users(X)*0.01)
8. Selective labeling:
If the fraud score is above a certain level, the case will enter a
queue for further investigation by human experts and the cases whose
fraud score are below are determined as clean by the human expert.
Decision making/Final Recommendation
The decision or the final Recommendation after analysis part is to
decide whether to ban the product or to trust the product. If the product
is banded by the admin then no user can view or buy the product
providing the user only the trustworthy products.
9. ANALYSIS
Existing System Proposed System
Simplifying access to information
is not done
Improves the productivity by
simplifying access to information
More time is required to decide
whether to trust the product or not.
Reduces the time to decide whether
to trust the product or not.
Involves Fraudulent Activities for
illegal profits
Fraudulent Activities are reduced
Delivers to the right person but not
always the good content
delivers the right content to the right
person to maximize immediate and
future business opportunities
10. DESIGN
Admin
User
Seller
Complaint filing
Fraud churn
The system can be broadly divided into the following modules:
11. • Login
• Authorize Sellers
• Manage sellers
• View complaints of the customer
• Decision to trust/block the products
An Admin performs the following actions :
This is represented in the following UML diagrams
ADMIN
The admin acts as an intermediator between seller and the
customer. An Admin is responsible to maintain the website information
giving a trust to the customers. If the admin feels all the products from
particular seller mostly are not trusted he can also remove the seller and
his related products.
12. Use case Diagram for Admin
Login
Logout
View Sellers
Admin
Manage Sellers
14. • Can add a new Product
• Can delete a product
• Can place New Offers to the product
• Can modify information related to the product such as
price ,basic information etc…
A Seller performs the following actions :
This is represented in the following UML diagrams
SELLER
The Seller module includes different sellers who wish to
sell their products. The seller needs to be approved by administrator
after a seller submits his registration. A Seller can add or delete or
modify information about different items.
15. Sequence Diagram for user
Login
Offers to Products
Logout
View Products
Seller
Edit information
16. • Register/Login
• View Products
• View Offers
• Purchase Products
• Give Complaint
A customer performs the following actions :
This is represented in the following UML diagrams
CUSTOMER
After successful registration, customer will be provided with a
gallery of different products which include the product name, Price, Sellers
name etc. While buying a product a customer can view the percent of
trustworthiness towards the product given by other users. After purchasing a
customer can also file complaint on that product where he feels
uncomfortable
17. Sequence Diagram for user Login
Login
View Products
Purchase Products
Logout
View Offers
Customer
file Complaint
18. databaseCustomer Gui validate userregister user
clicks on register
Enter details
user details
user created
save user
customer registered successfully
show message
login(usrnm,pwd)
validate userdetails
check user details
user details
validate user
user valid
login successful
19.
20. COMPLAINT FILING
• Buyers claim loss if they are recently deceived by fraudulent sellers.
• The Administrator views the complaints and the percentage of various
type complaints.
• Through complaints values the administrator set the trust ability of the
product as Untrusted or banned.
FRAUD CHURN
• Admin takes the decision whether to continue the seller to sell the
products or not.
• When some products are labeled as fraud by human experts, it is very
likely that the seller is not trustable and the products too.
• The fraudulent seller along with his/her cases will be removed from
the website immediately once detected.
23. int sold1=Integer.parseInt(sold);
int del1=Integer.parseInt(del);
int miss1=Integer.parseInt(miss);
int serv1=Integer.parseInt(serv);
int dam1=Integer.parseInt(dam);
int sum=del1+miss1+serv1+dam1;
Double sum1=sum/((0.01)*(sold1));
//System.out.println(sum1);
double t=50.0;
Double tru=100-sum1;
%>
28. CONCLUSION
We build online model for fraud product detection while concentrating
on customer needs. By empirical experiments on a real world online fraud
detection data, we show that our proposed online probit model framework,
which combines online feature selection, bounding coefficients from expert
knowledge and multiple instance learning, can significantly improve over
baselines . This can be easily extended to many other applications, such as
web spam detection, content optimization and so forth Websites that delivers
highly personalized and trusted experiences top the traffic and revenue
rankings across the globe.