カンファレンス (国内) ファイナンス掲示板情報からの株価予測
This paper describes the research on daily stock price prediction via Yahoo! Stock bulletin boards data. Though each words in bulletin board are treated as a feature for prediction in previous research, the words are often over-fitted because the same words are used in different cases. So we focus on the relationship of the word in this research. Concretely, we vectorize the words in bulletin boards with word2vec and we developed the estimation model with the vectorized words. As a result, it is confirmed that out proposed method performs well.
Paper : ファイナンス掲示板情報からの株価予測 （外部サイト）