Abstract:
Stock Market is the major influencer for national as well as international economy. There are
multiple factors which have key role in market movement involving the companies news and
performances, macro and micro economic factors, overall industry performance etc. The trend
of market can be measured from the driving factors or the past performance of stocks. Social
media has significant role in market analysis as investors and other social media users have
certain opinions about market trends. Stock market generally is not smooth and linear but Pak istan market specifically has lot of uncertainty and randomness. Sentiment analysis is a type of
opinion mining which we applied using twitter to get insights about future market. It is new
mode of machine learning for extracting opinion direction (negative, positive, neutral) from the
piece of text which is written for any entity like an organization, a person or a product etc. Inthis research, prediction performace was analyzed for four Oil & Gas marketing companies ofPakistan Stock Exchange. We have used the technique of TF-IDF for features extraction from
twitter data comprising the four companies. We are predicting the change price of market using
three models including Random Forest, Naïve Bayes and SVM. All of these methods are compared to get the best prediction model. The results depicted that Random Forest did better thanothers and achieved the accuracy of 53% for market prediction of all companies. Our future work involves the use of some fundamental observations, leveraging financial news and mathematical indicators to improve the accuracy.
