DETECTION OF FRAUD APP REVIEW USING SENTIMENT ANALYSIS AND MACHINE LEARNING

ABSTRACT:-

Abstract in today’s world ,the world is becoming smarter as the time passes . All it happens because of apps on the internet .With an increase in the range of mobile apps in day to day life, you should know which is fake and which are real . It’s based on a review of an app which is given by the user. Thus it’s necessary to keep track and develop a system to create a real app not a fake app. The target of this system is to develop a detecting fraud app before the user downloads it and harms the phone or system. The reviews might be fake or real. This will analyze the rating along with involvement of each user in the app. This system detects the fraud app on Google play store and App store before the installation of the app by the user on the phone. This system detects the review given by the user whether the review is positive, negative or neutral. Sometimes, reviews given by the user are fake. This system offers a positive review once the app is real or reviews get from the user to the current app is best. This system offers a negative review once the app is pretend or reviews get from the user to the current app is worse. This method offers a neutral review once the app isn’t real or pretend or reviews get from the user to the current app isn’t a lot of sensible or colossal. Sentimental analysis is used in this detecting system. Sentiment is a feeling or perspective prompted by the emotions of the client. Sentiment analysis is additionally named as opinion mining, as the opinion area unit collected from the client is deep-mined to reveal the rating of the app.

Keywords: Mobile app, Review, Rating, Fake or real app, Positive, negative or neutral, Sentimental Analysis, SVM, Random forest, Decision tree, Naïve bayes

OBJECTIVES OF THE PROJECT :-

The objectives of the Detection of fraud apps using sentiment analysis are:

To design a system which may detect fake app review by considering different evidence indicating their true behavior.
To find app reviews are real or not.
To increase the classification accuracy of a system.
To find the review of an app is positive, negative or neutral.
To find an app review is real or fake from the emotions of the client.
To develop such a system that notices reviews based mostly on evidence thus aggregation supported optimization to mix the evidence for detection of fraud.
To demonstrate the Framework of Fraud positioning revelation in portable applications based on review.

EXISTING SYSTEM:-

In this paper, they know fraud apps by exploitation apps and sentiment analysis. They introduced FairPlay, a novel system that discovers and profits traces left behind by fraudsters, to find each malware and fraud apps subjected to search rank fraud. FairPlay correlates review activities and clearly combines detected review relations with linguistic and behavioural signals from Google Play app knowledge so as to spot suspicious apps. They show that 75% of the known fraud apps interact in search rank fraud.Users square measure dispute into writing positive reviews, and install and review different apps..

Android applications are widely utilized by countless users to perform many various activities. However, several applications have been according to fake behaviour not matching with their expected behaviour. The present relevant approaches that establish these applications suffer from performance problems wherever the spread of false negatives stay high. These approaches are not scalable and provide very little flexibility to question supported groups of suspicious permission to spot a group of extremely relevant known abnormal applications. The known most relevant application category’s permission is then checked to seek out if there’s important overlapping or to not establish associate application as suspected anomalous. We have a tendency to apply Latent linguistics classification to spot malware applications. Their initial analysis results counsel that the planned approach will establish malware applications accurately. This work planned a replacement approach to find abnormal humanoid applications by characteristic the foremost relevant class of permissions to match from and so confirming behaviors in associate ape atmosphere. The relevant class is known using Latent linguistics Index (LSI) analysis. The approach has the potential to get new abnormal applications. We evaluate the approach exploitation 2 open supply malware repositories. The initial results show that our approach will find malware applications. Our future work includes evaluating additional applications and question varieties, and applying the approach to address different security vulnerabilities.

PROPOSED SYSTEM:-

With the growth within the quantity of web framework, to differentiate the false statement app review, this undertaking proposes a simple and prosperous framework. Consequently, it’s tough to identify why people like or dislike a specific app. Sometimes, users give the fake review to get a worse rating to the app. We aim to unravel this drawback. We demonstrate the Framework of Fraud positioning revelation in portable applications. Thus we are proposing associate humanoid applications which can method the knowledge, comments and three reviews of the appliance with communication process to present results.

Detection of fraud apps is based on sentiment analysis and user emotions. Sentiment analysis handles the emotions of the client.If the result of this analysis is positive,then the app is good for use. If the result of this analysis is negative,then the app is a false statement app. If the result of this analysis is neutral,then the app is not much good or not much bad for use. The main aim is to develop such a system that notices reviews based mostly on evidence thus aggregation supported optimization to mix the evidence for detection of fraud. Therefore it’s going to be easier to detect the fraud applications. As a developer, it’s essential to “stay on top of your game”, i.e., keep your app updated with the most requested features and bug-fixes. However, most app stores give only an average rating (out of 5) for every app.

MODULES:-

User Register: User have to register to check whether app review is positive, negative or neutral.
User Login: User have to register to check whether app review is positive, negative or neutral.
Processing Sentence: Use Natural language processing (NLP) to makes possible for computers to read text, hear speech, interpret it, measure sentiment and determine which parts are important.
Word to Vector: Tf-idf is used to transforming text into a numerical feature is called text vectorization which mathematically eliminates naturally occurring words in the English language, and selects words that are more descriptive of your text.
Classifying Sentence: Based on Tf-idf vectors machine learning algorithms (SVM, Naive bayes, Random Forest, Decision tree) will classify text.

ADVANTAGES :-

The proposed framework is scalable and can be extended with other domain generated evidence for ranking fraud review detection.
To the best of our knowledge, there is no existing benchmark to decide which leading sessions or Apps really contain ranking fraud. Thus, we develop four intuitive baselines and invite five human evaluators to validate.
Experimental results show the effectiveness of the proposed system, the scalability of the detection algorithm as well as some regularity of review fraud activities.
It keeps your app updated with most requested features and bug free based on user review.

HARDWARE AND SOFTWARE REQUIREMENTS

HARDWARE:

Processor: Intel Core i3 or more.
RAM: 4GB or more.
Hard disk: 250 GB or more.

SOFTWARE:

Operating System : Windows 10, 7, 8.
Python
Anaconda
Spyder, Jupyter notebook, Flask.
MYSQL