Combining Sentiment Analysis with Socialization Bias in Social Networks for Stock Market Trend Prediction
Abstract
According to the indirect relationship between information and stock trend, information such as comments and tweets can be used for stock trend prediction. When conducting classification on text data, feature sparse issues occur during conversion between tweets and word vectors. Another problem is that the unreliability of average sentiment scores to indicate one day’s sentiment. This is especially caused by the unbalanced number between positive and negative within one day, thus a large bias between sentiment and stock trend arises. In addion, information has social attributes when created and diffused in social networks, bias containing people’s belief in social networks also have become socialization bias. In order to solve those problems, this work proposes a sentiment analysis based prediction model and an inverse bias algorithm. Instead of applying sentiment analysis to add sentiment related features, this work uses SentiWordNet to give an additional weight to the selected features, and applies two kinds of sentiment analysis to inverse the socialization bias. Aiming at labeling the tweets to sentiment related groups to help find socialization bias, this work also proposes an extended wordlist based on a semi-supervised Naïve Bayes classification algorithm. After finishing the inverse socialization bias, stock trends were used to label example sets. Different classification algorithms were compared in this work. The proposed model with SVM linear algorithm proves to yield accuracy of 90.33% at its best performance.
Remember to check out the Most Cited Articles! |
---|
Check out these titles in artificial intelligence! |