A Time Series Analysis of Trends With Twitter Hashtags Using LSTM

Social media has converged into our everyday life in such a way that from lifestyle to our behaviors, we follow the social media immensely. The influence of social media is easily observable whenever any ‘TREND’ occurs in social media platforms and we stumble over the internet to follow that. Detecting trends can help to get notified about not only the ongoing topics around the internet as well as it gives us the chance to understand people's choices, emotions and so on. This study of trend analysis becomes more specific and invaluable when it is targeted for a specific genre or community. As such- Gaming, Movies and Tv series viewers, etc. There are very few assets in twitter for those specific genres which could assist its audiences or consumers to keep track of the trend list of that particular genre. Our chosen genre was Gaming. The primary purpose of our research is to analyze and predict the ongoing trends around Twitter of this specific community using Twitter Hashtags, which is a short yet quite stronger mode of expressing one's mood or the gist of any topic. Our contribution to the research reflects in applying the LSTM model of Recurrent Neural Network model in a time series dataset which is more convoluted and less explored area of research than the existing methods of analyzing regular time series models.

I. Introduction

Social media and its trends are shaping our daily life even as we speak. These concurrent issues and topics have been a major catalyst in our everyday lifestyle. From fashion to our emotions, the impact of trendy topics is noticeable. Founded in 2006, twitter has become very popular quite rapidly for its unique feature. The reason for choosing twitter and the hashtags for the research was that they are short and specific yet highly thought-proviking. But the inadequacy of specific information on a particular genre on Twitter encouraged the researchers to conduct this research. For instance, if anybody wants to see or observe which games or movies are currently on the trendy list, there are no such data for him to see. The hashtags in the trendy list are on so much vast and wide issues that anybody who is looking for particular information on for any particular community, will be disappointed. Thus, the research model is built in such a way that will help any particular community to find what is on the trendy list for that community. For instance, if a movie lover decides to observe which movies people are talking about now and will be talking about for the next few days, our system will be able to track that for him. For research, the gaming genre was chosen. That's because besides recreation, the gaming industry is quite big as well as profitable. In twitter, there is not much information to obtain the trendy games list for the gamers. The primary motivation for the research was to observe a factor that is closely related and impactful for human behavior that is trendy topics. Furthermore, to specify the research, working on a micro-level of a popular social media seemed to be a smart option that is Twitter hashtags. Finally, the research was conducted for particular communities so that they can predict the future trendy list of the issues of their communities.