FILTERING FAKE PRODUCT REVIEWS BY USING n GRAM APPROACH
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
Mukt Shabd Journal ISSN NO : 2347-3150
FILTERING FAKE PRODUCT REVIEWS
BY USING n GRAM APPROACH
Ms.N.Padmapriya,
Assistant Professor,
Department of Computer Science and Engineering, SSM Institute of Engineering and Technology, Anna
University,Dindigul, Tamil Nadu
A.Soundharya,
UG Scholar
Department of Computer Science and Engineering, SSM Institute of Engineering and Technology, Anna
University,Dindigul, Tamil Nadu
V.Subalakshmi,
UG Scholar
Department of Computer Science and Engineering, SSM Institute of Engineering and Technology, Anna
University,Dindigul, Tamil Nadu
R.Sugapriya
UG Scholar
Department of Computer Science and Engineering, SSM Institute of Engineering and Technology, Anna
University,Dindigul, Tamil Nadu
Abstract--- As the trend to shop online is increasing day by day and more people are interested in buying the
products of their needs from the online stores. This type of shopping reduces the shopping time and travelling
time of the customers. Customers go to online store, search the item of their need and place the order. But, the
thing by which people face difficulty in buying the products from online store leads to bad quality of the
product. Customer place the order only by looking at the rating and by reading the reviews related to the
particular product. Such comments of other people are the source of satisfaction for the new product buyer.
Here, it may be possible that the single negative review changes the angle of the customer not to buy that
product. In this situation, it might possible that this one fake review makes the loss and business. So, in order
to remove this type of fake reviews and provide the users with the original reviews and rating related to the
products, we proposed a Fake Product Review Monitoring and Removal System (FaRMS) which is an
Intelligent Interface and takes the Uniform Resource Locator (URL) related to products of Amazon, Flipkart
and Daraz and analyzes the reviews, and provides the customer with the original rating. It is a unique quality
of the proposed system that it works with the three e-commerce Websites and not only analyzes the reviews in
English but also the reviews written in Tamil and Hindi. The proposed work achieved the accuracy of 87% in
detecting fake reviews, of written in English by using intelligent learning techniques which is greater than the
accuracy of the previous systems.
Volume IX, Issue VII, JULY/2020 Page No : 2132Mukt Shabd Journal ISSN NO : 2347-3150
1.INTRODUCTION
There are different ways to shop like people can buy a specific thing of their need by going to a store or mall. In this
style of shopping the seller gives the feedback of the product, they do not know whether he/she is giving a fake
feedback or original. Because, it is upon seller honesty, how much the seller is true in his/her words and people have
to carefully examine the product because people do not have any
other option in examining the product. If customer do not pay attention in buying that product then it may be proved
a waste for them. On the other hand, nowadays source of shopping has been changed. People can buy the products
from the online stores of different brands. Here, customer have to place the order without seeing and examining the
original product. Every customer reads the reviews and buy the product. Therefore, we are dependent on the reviews
about the product. These reviews may be the original or fake. The customer wants to buy an original and reliable
product, which is possible only when you get the original feedback related to that product. can find the original
feedback and rating related to a specific product .Then, it is the source of satisfaction and reliability for customer. In
the proposed technique, the reviews related to a product for which the URL is given are extracted. After it, the system
finds the fake reviews and finally by analyzing these reviews system finds the original reviews of the product. Previous
researches detect fake reviews using different approaches including identification address, opinion mining and
sentiment analysis, machine-learning approach. There are many researches available (English, Tamil, Hindi ) detect
fake reviews for Therefore, we have proposed Fake Product Review Monitoring and Removal System (FaRMS) in
which a customer can get the best possible item from the online store in a short time and with the original reviews
associated with that product. This system gives you the original words of people related to the product with genuine
reviews. Some popular products can get hundreds of reviews at some large merchant sites and FaRMS gives you the
promising reviews by filtering fake reviews and then customer can decide whether they want to buy or not. Fake
review monitoring system focuses on detecting spam and fake reviews by using sentimental analysis removes the
reviews which have curse and vulgar words. In the proposed system web crawler is used to scrap the data on the
Website. In the preprocessing, the data is converted into the required format and then the fake reviews are removed
from the mixture of original and spam reviews. Fake reviews are detected by the Fake Review Detector.
2.LITERATURE SURVEY
Anjay K S., [1] In this paper, they proposed a framework to detect fake product reviews or spam reviews by using
Opinion Mining. The Opinion mining is otherwise called as Sentiment Analysis. In sentiment analysis, they try to
figure out the opinion of a customer through a piece of text. They first take the review and check if the review is
related to the specific product with the help of Decision tree and use Spam dictionary to identify the spam words in
the reviews. In Text Mining they apply several algorithms and on the basis of these algorithms we get the specific
results.
Rakibul Hassan., [2] Decision making on online products for purchase was mostly depends on reviews given by the
users. Hence, for their own interests, opportunistic individuals or groups try to manipulate the product reviews. To
detect fake online reviews, this paper introduces some semi-supervised and supervised text mining models as well as
it compares the efficiency of both techniques on dataset containing hotel reviews.
Volume IX, Issue VII, JULY/2020 Page No : 2133Mukt Shabd Journal ISSN NO : 2347-3150
Nidhi A. Patel., [3] Based on reviews or opinions that are written by others based on their experiences helps the
customers to take decision directly. In this competitive world any person can write anything, this raise the number of
fake reviews. Various companies are hiring people to write fake positive reviews about their services or products or
unfair negative reviews about their competitors' services or products. Hence we need a system to detect such fake
reviews and remove them. In this paper we discuss various supervised, unsupervised and semi supervised data mining
techniques for fake review detection based on different features.
Jiandun Li., [4] In this paper, how to distinguish whether a review is fraudulent or a reviewer is a spammer has long
been studied, but the question of general review pattern mining is still open. In this paper , they model online product
review systems into bipartite networks and adopt a network technique, called the weighted motif to uncover underlying
reviewing patterns.
Anusha Prabakaran., [5] In this paper, a statistical credibility scoring mechanism to identify spam reviews. It consists
of three components: detection of duplicate reviews, detection of anomaly in review count and rating distribution, and
detection of incentivized reviews. These three methodologies complement each other to effectively indicate the
credibility of product reviews without requiring significant computational resources. It can aid data mining and online
spam filtering systems to filter out spam product reviews and refine product rankings.
3. PROPOSED SYSTEM
In this proposed system, the sentences that are not related with the quality of a product such as customer service or
sentence related to the. In this paper the preprocessing is done by Support Vector Machine (SVM). First of all it
removes the comments which neither is nor related with the quality of the product. Second stage describes the weights
of the reviews based on the votes. The final stage calculates the overall ranking of the product. The ranking score is
calculated by the relevance of the review with quality of the product, review content, and posting date of the review.
In the evaluation process they use two measures to quantify effectiveness of the ranking model which are as following:
correlation between the ranking method and the Amazon’s rank and second is the Mean Average Precision (MAP),
which is a very commonly used technique for Evaluating ranking accuracy. As this system is finding the fake reviews
by using the only two properties of the reviews but as per the future work describes in the paper more properties can
be used to find out the fake reviews more accurately.
Volume IX, Issue VII, JULY/2020 Page No : 2134Mukt Shabd Journal ISSN NO : 2347-3150
Reviews
Data Pre processing
Review Analysis
Sentiment
Classification
Results
Figure 1 System Architecture
Module Description
Admin : This module describes about the admin of this project. Here administration is the area of the Web store that
is accessed by the merchant to manage the online shop. In general a store manager is able to add and edit products,
categories, shipping and payment etc. Admin is the person who maintains the whole system. Admin updates the new
products of the branded company.
User Registration and Login : To search and rate any product in this application all user must registered with the
system. User must register with this application for access the product list uploaded by the admin. User registers in
this application with their details such as Name, gender, age, mobile number, email, username and password. After
successful registration user able to login into the system then search for a product and rate any product based on the
satisfaction they realized in that product.
Volume IX, Issue VII, JULY/2020 Page No : 2135Mukt Shabd Journal ISSN NO : 2347-3150
Product Upload : In this application admin can only add new product details to search and rated by user. Admin need
to login into this system by providing his/her username and password, after successful login admin can add the new
product details such as product name, product ID, Quantity of product and cost, description and the image of the
particular product. These details are important to search by the user. The products are search by their name and/or
which category the product belongs to.
User Rating : After successful login user can search for a particular product and rate the product. The rating is based
on the 5 Star Rating. If the user give 5 star to a product it’s 100% satisfied the user, and if any user give 1 Star for a
product then it’s not up to user’s expectations. Based on the user rating the product is recommended to other user in
the same category or in same age group. This rating is very important for the system to maintain the high rated and
mostly liked products and recommend it to other user of this application.
Recommendation : Three kinds of methods were proposed to compute better recommendation algorithms. They are
instance selection, time window and ensemble learning.A set of dynamic features are used to describe users multi-
phase preferences.
Annotate reviews
Yes
Review Acceptab
collection le
No
Calculate Accuracy
Dataset
Partitioning
Run on Test Data sets
Sentiment
Analysis
Build Model
Map sentiment
with trained Classify using
data having identified vectors
fakeness labels
Modify Vectors
Volume IX, Issue VII, JULY/2020 Page No : 2136Mukt Shabd Journal ISSN NO : 2347-3150
Figure 2 Module Interface Diagram
Fake view Detection : Some challenging e-shopping website can give fake feedback to the original website through
some persons. To achieve popularity in internet marketing, these activities can be done. To find the fake review of the
products user’s IP address are stored at every product rating. Rating rate will be cancel if particular user’s IP address
repeated with different rating for the same product. This system automatically gives the notification .
4. Results and Discusssion
In this chapter , we discuss the various output results.
Home Page :
Admin Login :
Add Product :
Volume IX, Issue VII, JULY/2020 Page No : 2137Mukt Shabd Journal ISSN NO : 2347-3150
View Product :
Customer View :
Purchase Report :
Rating and Suffusions :
Volume IX, Issue VII, JULY/2020 Page No : 2138Mukt Shabd Journal ISSN NO : 2347-3150
User Registration :
User Login :
Product View :
Display the detail :
Volume IX, Issue VII, JULY/2020 Page No : 2139Mukt Shabd Journal ISSN NO : 2347-3150
Rating Suggestion :
Purpose :
Rating :
Review the Products :
Volume IX, Issue VII, JULY/2020 Page No : 2140Mukt Shabd Journal ISSN NO : 2347-3150
Fake Review Message :
Reviewing Turn over Messages :
5. Conclusion
In the proposed work, dataset is developed that contains Urdu and Roman Urdu reviews. It is difficult to detect fake
reviews by yourself. So, n-gram approach is used to detect fake reviews for multiple languages. It is observed that the
text categorization with SVM classifier is best approach for the detection of fake reviews. Now a days, as the
technology is growing day by day and there are so many Websites and applications available in the online market by
which seller can sell their products and on that products there are millions of reviews available. There are some
organizations posting fake reviews for the products of the seller in order to increase or decrease the rating of the
products. Therefore, the system is proposed that detects the fake reviews in multiple languages including English,
tamil, and Hindi, classify the reviews in genuine.
Volume IX, Issue VII, JULY/2020 Page No : 2141Mukt Shabd Journal ISSN NO : 2347-3150
References
[1] A. Sinha, N. Arora, S. Singh, M. Cheema, and A. Nazir, “Fake Product Review Monitoring Using Opinion
Mining,” vol. 119, no. 12, pp. 13203– 13209, 2018.
[2] Torbet, Georgina. “U.S. Customers Spent over $6 Billion on Black Friday Purchases.” Digital Trends, Digital
Trends, 25 Nov. 2018, www.digitaltrends.com/ web/shopping-totals-blackfriday/.
[3] Sterling, Greg. “Study Finds 61 Percent of Electronics Reviews on Amazon Are 'Fake'.” Marketing Land, 19 Dec.
2018, marketingland.com/study-finds-61-percent-of electronics-reviews-on-amazon-are-fake 254055.
[4] K. Khan, W. Khan,A. Rehman, A. Khan, Asfandyar. Khan, A. Ullah Khan, B. Saqia, "Urdu Sentiment Analysis,"
(IJACSA) International Journal of Advanced Computer Science and Applications, Vol. 9, No. 9, 2018.
[5] A. Mukherjee, B. Liu, and N. Glance, "Spotting Fake Reviewer Groups in Consumer Reviews," 2012.
[6] A. Mukherjee, V. Venkataraman, B. Liu, and N. Glance, “What Yelp Fake Review Filter Might Be Doing?,” Aaai,
pp. 409–418, 2013.
[7] Z. Wang, Y. Zhang, and T. Qian, “Fake Review Detection on Yelp Dataset and features,” pp. 1–6.
[8] S. Xie, G. Wang, S. Lin, and P. S. Yu, “Review spam detection via temporal pattern discovery,” p. 823, 2012.
[9] C. Paper, “Mining millions of reviews : A technique to rank products based on importance of reviews Mining
Millions of Reviews : A Technique to Rank products Based on Importance of Reviews,” no. November, 2015.
[10] V. K. Madhura N Hegde, Sanjeetha K Shetty, Sheikh Mohammed Anas, “Fake product review monitoring,” Int.
Res. J. Eng. Technol., vol. 05, no.06, p. 4, 2018.
Volume IX, Issue VII, JULY/2020 Page No : 2142You can also read
