Social Media and Sentiment Analysis of Nifty 50 Index

Abstract

Now-a-days, micro blogging social media platforms such as twitter, foursquare, ello etc. are used extensively by a diverse range of users to express their explicit opinions about a diverse range of topics on the internet. This paper aims to collect this freely available data using web scraping techniques and analyse it to assess the general sentiment prevailing in the market regarding the Nifty 50 index. Opinions are collected explicitly from twitter using the twitter application program interface (APl). A total of 154,398 tweets were collected and analysed to calculate the proportion of positive sentiment in the markets between Jan. 1, 2016 and Dec. 31, 2017. For assessing the sentiment of a given sentence, SentiWordNet, a lexical resource for opinion mining, was used. The average sentiment for all the tweets is calculated to assess the general sentiment prevailing in the market about the Nifty 50 index on a particular day. To understand the relationship between both the variables, linear regression analysis is performed, whereby, natural log of Nifty 50 index has been taken as the dependent variable and 10 days Moving Average (10MA) of proportion of positive sentiments as the independent variable. Based on the regression analysis, the coefficient of regression was obtained as -0.602 and its p-value was obtained as 0.00, indicating that there is a significant relationship between the 10MA of proportion of positive sentiments and natural log of Nifty 50 index. As the p-value of the regression is significant, it gives us a scope to do further analysis on the collected data. If significant, traders can have an additional tool in their technical analysis kit.