Wallet App Credibility Analysis Based on App Content and User Reaction

WalletApp Credibility Analysis Based on App Content and User Reaction

Usage of online banking and mobile banking has already facilitated and added a dramatic change in customer’s life and now mobile wallets have made it easier by holding money digitally. According to reports, market leader PayTM in this space is downloaded by 40% of smart phone users in India and it has grown from 150-180 million-user base in two months [1]. According to Reserve Bank of India (RBI), a massive growth lies in the digital wallet adoption and usage as prime mode of payment [2]. RBI stated following points related to mobile wallet. In year 2015-2016, mobile wallet transactions were more than 490 billion, which was approximately five times more than the year 2014-15.

https://codeshoppy.com/android-projects-titles-ieee.html

A year-wise wallet app growth chart published by Reserve bank of India is shown in Figure 1 (a). Transactions through mobile wallets have grown 500% in between year 2014-16 (see Figure 1(b)). Digital wallet facilitates customers to digitally pay for a myriad of services but substantial utilization depends on reliability and acceptance of wallet app by its users. 

Most important factors to validate and analyze reliability and market acceptance of a wallet app are reviews and ratings as user-generated content. Surely, more reviews and ratings on an app give better probability to be explored and discovered to get knowledge out of it. Over half of people read a minimum of one review before they decide to download an application [3]. Number of real scenario cases exists in which fraudsters deploy number of malicious rating and reviews to apps, disguised as a reliable application. These fraudulent behaviors of fraudsters dissuade the users to download the app. This activity refers as Opinion Spam or shilling attack in technical language that helps to mislead users by providing undeserving positive reviews to raise popularity of an app and negative reviews to damage reputation of other competitor apps [7]. Reaction

 According to Amoroso and Watanabe’s [11] model for Mobile Wallet Consumer Adoption, social influence and trust are key factors towards increase in usage of an e-wallet. This is where user reviews come into play, they shape a person’s initial mind-set towards the app. Trends involving individual fake reviews such as review length, reviewer deviation, maximum number of reviews etc. Mukherjee et. al. has discussed in [5] the analysis of Yelp fake review filter by using supervised learning with the help of feature sets comprising of word bigrams and unigrams.
Another study [6] identified trends involving group spammers and stressed that these groups can do a lot more damage than individual spammers can. This study uses frequent item set mining and labels them to measure spamicity and build both group and individual spamming factors such as group time window, group member content similarity, individual content similarity and individual rating deviation. Majority of the opinion spamming studies are performed for sentiment analysis of the content. Usage of SentiWordNet [5] is one such approach, wherein SentiWordNet, a predefined lexicon dataset has been used to filter the positive and negative words in review dataset with a remarkable accuracy. 

Algorithms such as K-Means [9, 10] are basically used for clustering the data, which was later fed to SVM for classification.With breathtaking advances and wide adoption of wallet app, while providing valuable digital payment utility and convenience to user, also bring new issues and evaluation focus (EF) specifically for wallet apps in user’s mind. Some evaluation focus points for wallet app which are handled in our work are as follows- EF1: Which Wallet App provides more features?EF2: Which is having more number of positive / negative reviews? EF3: Which contains more feature-oriented reviews?EF4: Do we make some association in wallet review based rating and user provided App rating?In this paper, we attempt to analyze credibility of wallet application, by incorporating score to an application based on app description and user reactions. 

Certainly, reviews have great role in application downloads. More number of good reviews and ratings on an app enhance user’s interest towards that specific app. On these lines, we frame our line of study that reviews should qualify the application rating. 

In our work, we try to find out wallet apps providing maximum number of features with the help of their description analysis, detect app those having maximum number of users’ discussion regarding their features and finally match the app rating published with the credibility score computing using proposed credibility analysis process for all wallet apps. 
Wallet applications those having difference in app rating and computed normalized app score higher than a threshold value are considered to be untrustworthy application. Here, we classified wallet apps in three categories - untrustworthy wallet applications, opinion spam applications and reliable wallet applications.