Stock Price Forecasting with Support Vector Regression Based on Social Network Sentiment Analysis and Technicl Analysis

This study predicts Price of stocks in the short term by using the analysis of investors' opinions of the social network. The predictability of stock markets, due to having a complex, dynamic and nonlinear system that it has always been one of the challenges for researchers. The effect of users' feelings on the social network and its combination with 20 technical indicators on the accuracy of stock price forecasting. The study period is from the beginning of April 2016 to the end of March 2017 (two years). To access sufficient data, a sample of 14 active companies that had the most comments and posts. Data mining of technical indicators was performed and support vector regression was used to predict. The results show that the use of technical indicators is more accurate compared to combining it with the aggregation of users' emotions and has less RMSE errors. The number of comments has a significant correlation and the results of Granger causality test showed that it is possible to use the aggregation of users' daily emotions to predict stock prices.