A Real-Time Twitter Trend Analysis and Visualization Framework

Introduction

Twitter is considered as one of the world’s largest social networking sites which allow users to customize their public profile, connect with others and interact with connected users. A global survey from one of the well-known companies like “Statista” witness that, as of the fourth quarter of 2017 there are average number of 330 million monthly active twitter users (Statista, 2017). Users post short messages of 140 characters called tweets based on variety of topics ranging from simple ones such as “Hi #xyz☺am@college” to themes such as “#IPL-2017” using different ways such as blogging on Twitter website, using Twitter mobile application and also through other social network applications which allow virtual connections from one application to another such as Instagram. Being a most popular social networking sites twitter uses a network model called “following”. Any person can follow any other person who remains to be his/her friend. The person who follows is named as a follower and a follower on twitter can receive all the updates that he/she follows. Tweets are of different types which include normal tweet, reply and retweet. Normal tweets are the one which people post based on their thoughts and opinions which is not a reply to any others tweets or a retweet for concerned tweet. Reply is a tweet which a twitter user post with “@” symbol attached with name of replies’ for example “@abc am not interested in you”. Retweet is a message which commonly starts with “RT” and it is a post which is shared by the follower to his/her followers. Apart from this the other very important features of tweet are HashTags and URLs which make twitter stream more readable and provides an understanding of current topic or an event discussed on twitter (Lau, Collier & Baldwin, 2012).

Twitter has outcome as one of the best platforms of news information propagation in present days due to enormous number of users’ active on daily basis and also, it’s very spontaneous nature of real-time tweeting that reach out to people within few seconds. It may be any kind of news such as, Prime Minister’s decisions of changes in autonomy or company’s date of releasing their product, the “following” feature in twitter makes the news to spread around within short interval of time. Nearly there are thousands of topics discussed on twitter daily but most of the users are interested in only the hot topics or top news which is going viral around the world and will remain as trend in next few hours. Due to enormous number of tweets on twitter it is impossible for us to browse the top news topic or trends. Hence there is a need for a real-time trend analysis system which can analyse and predict the top trends which are going viral around the world and will remain to be trend in next few hours or in next few days. The existing systems for trend analysis used traditional techniques such as LDA topic Modelling, TF-IDF algorithm etc for predicting top trends on twitter. But these methods are less accurate, and the results were satisfactory. Thus, for analysing and predicting top trend on twitter a real-time twitter trend analysis and visualization framework is introduced in this research work by implementing a novel method called “Twitter Trend Momentum (TTM)”. A deep reason analysis for trend variations is carried out as a major part of this research work and finally the framework is applied to build an interactive web application called “TwitTrend” which acts a real-time update and recommendation system for trend detection and analysis.