Abstract: Sentiment  Analysis  is  an  innovative  development  technique  that  uses  natural language processing techniques to derive people's emotions under positive, negative,and neutral based on public opinions of information. The main objective of this study is  to  introduce  a  novel  stock  market  prediction  method  based  on  the  Grey Exponential Smoothing method for analyzing social media data within a big-data distributed environment. The empirical investigation of this study is mainly carried out based on the stock market price indices parallel to the extracted Tweets collected during the three selected politically important moments that happened in Sri Lanka during the past ten years; the first case study is based on the political background after  the  ending  of  the  thirty  years  of  civil  war  in  years  2009.  In  the  year  2015,Maithripala Sirisena ended the dynastic rule of Mahinda Rajapaksa. So, the second case  study  has  based  the Tweets  on  the  political  reforms  done  after  the  2015 presidential  election;  the  third  study  is  based  on  the  Sri  Lankan  political  and economic  background  after  the  Rajapaksas  rose  again  in  2020.  For  validations purpose, K Nearest Neighbour, Decision Tree Model, Support Vector Machine, Grey Exponential Smoothing model, and Multinomial Naïve Bayes machine learning were considered.  According  to  the  empirical  findings,  the  new  proposed  Hybrid  Grey Exponential Smoothing model is highly accurate with the lowest RMSE error values in one-head forecasting. Furthermore, the key finding of this research suggested that the  hybrid  Grey  Exponential  Smoothing  model  performs  well  in  sentiment classification-based financial predictions than traditional methods, especially with non-stationary behavioral backgrounds. 

Key words: Sentiment Analysis, Grey Exponential Smoothing Model, Time Series Analysis