Abstract: - :- Online Modeling: - :-
We
consider the problem of building online machine-learned models for detecting
auction frauds in e-commerce web sites. Since the emergence of the World Wide
Web, online shopping and online auction have gained more and more popularity.
While people are enjoying the benefits from online trading, criminals are also
taking advantages to conduct fraudulent activities against honest parties to
obtain illegal profit.
Hence
proactive fraud-detection moderation systems are commonly applied in practice
to detect and prevent such illegal and fraud activities. Machine-learned
models, especially those that are learned online, are able to catch frauds more
efficiently and quickly than human-tuned rule-based systems. In this paper, we
propose an online probity 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 online auction fraud detection data we
show that this model can potentially detect more frauds and significantly
reduce customer complaints compared to several baseline models and the
human-tuned rule-based system.
Existing System
The
traditional online shopping business model allows sellers to sell a product or
service at a preset price, where buyers can choose to purchase if they find it
to be a good deal. Online auction however is a different business model by
which items are sold through price bidding. There is often a starting price
and expiration time specified by the
sellers. Once the auction starts, potential buyers bid against each other, and
the winner gets the item with their highest winning bid.
Proposed System
We
propose an online probity 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 online auction
fraud detection data we show that this model can potentially detect more frauds
and significantly reduce customer complaints compared to several baseline
models and the human-tuned rule-based system.
Human
experts with years of experience created many rules to detect whether a user is
fraud or not. If the fraud score is above a certain threshold, the case will
enter a queue for further investigation by human experts. Once it is reviewed,
the final result will be labeled as Boolean, i.e. fraud or clean. Cases with
higher scores have higher priorities in the queue to be reviewed. The cases
whose fraud score are below the threshold are determined as clean by the system
without any human judgment.
Module Description
Rule-Based Features
Human
experts with years of experience created many rules to detect whether a user is
fraud or not. An example of such rules is “blacklist”, i.e. whether the user
has been detected or complained as fraud before. Each rule can be regarded as a
binary feature that indicates the fraud likeliness.
Selective Labeling
If
the fraud score is above a certain threshold, the case will enter a queue for
further investigation by human experts. Once it is reviewed, the final result
will be labeled as Boolean, i.e. fraud or clean. Cases with higher scores have
higher priorities in the queue to be reviewed. The cases whose fraud Score are
below the threshold are determined as clean by the system without any human
judgment.
Fraud Churn
Once
one case is labeled as fraud by human experts, it is very likely that the
seller is not trustable and may be also selling other frauds; hence all the
items submitted by the same seller are labeled as fraud too. The fraudulent seller along with his/her cases
will be removed from the website immediately once detected.
User Complaint
Buyers
can file complaints to claim loss if they are recently deceived by fraudulent
sellers. The Administrator views the various type of complaints and the
percentage of various type complaints. The complaints values of a products
increase some threshold value the administrator set the trust ability of the
product as Untrusted or banded. If the products set as banaded, the user cannot
view the products in the website.
System Configuration
Hardware Configuration
·
Processor -
Pentium –III
·
Speed - 1.1
GHz
·
RAM - 256 MB (min)
·
Hard Disk - 20
GB
·
Floppy Drive -
1.44 MB
·
Key Board -
Standard Windows Keyboard
·
Mouse - Two
or Three Button Mouse
·
Monitor -
SVGA
Software Configuration
·
Operating System :
Windows95/98/2000/XP
·
Application Server :Tomcat5.0/6.X
·
Front End : HTML, Java, Jsp
·
Scripts : JavaScript.
·
Server side Script : Java Server Pages.
·
Database : Mysql
·
Database Connectivity :
JDBC.
Conclusion
In
this paper we build online models for the auction fraud moderation and
detection system designed for a major Asian online auction website. By
empirical experiments on a real world online auction fraud detection data, we
show that our proposed online probity model framework, which combines online
feature selection, bounding coefficients from expert knowledge and multiple
instance learning, can significantly improve over baselines and the human-tuned
model.
Note
that this online modeling framework can be easily extended to many other
applications, such as web spam detection, content optimization and so forth.
Regarding to future work, one direction is to include the adjustment of the
selection bias in the online model training process. It has been proven to be
very effective for offline models in [38]. The main idea there is to assume all
the unlabeled samples have response equal to 0 with a very small weight. Since
the unlabeled samples are obtained from an effective moderation system, it is
reasonable to assume that with high probabilities they are non-fraud. Another
future work is to deploy the online models described in this paper to the real
production system, and also other applications.
0 comments :
Post a Comment