Machine Learning Based Prediction of Stock Exchange on NASDAQ 100: A Twitter Mining Approach

Twitter is a popular social media network. Its impact on our daily life is increasing day by day. Twitter is not only used for current affairs but also for showing its significance in corporate culture. The motivation for this research is to elaborate on the importance of user tweets on famous companies around the world by analyzing the user tweets according to their sentiments, i.e., optimistic (positive), harmful (negative) and disinterested (neutral). This analysis is an effort to show the corelation between user tweets and the variation in the stock market share prices of companies worldwide. The dataset was taken from the NASDAQ 100, the famous source of stock price data from world-famous companies. After data extraction, a sample from the entire dataset was used to obtain a minimum but optimal data subset. One-Hot Encoding technique has been used for features reduction to reduce the useless data and make the data more flexible to use. For modelling, the SVM machine learning algorithm is used with different kernels. Furthermore, cross-validation is leveraged to evaluate and verify the model's accuracy and stability. The maximum 92% accuracy was achieved using SVM with Linear kernel, whereas RBF kernel accuracy was 62%, which was low compared to the former one.)