株価に影響を与える重要な出来事が記載された記事の自動抽出

Abstract

This paper proposes a method of extracting important articles that influence the stock price of companies from financial articles. In the first step, our method obtains dates that have large difference from previous day at stock price of a selected company. In the second step, our method acquires articles that are published around the obtained dates and concern the selected company from financial articles. In the third step, our method extracts articles that influence the stock price of the selected company by using SVM as a machine learning method from the acquired articles. Finally, we evaluated our method. As a result, our method achieved 69.8% precision and 53.1% recall.