Rachlin+07

テキスト分析

Gil Rachlin, Mark Last, Dima Alberg, Abraham Kandel: ADMIRAL: A Data Mining Based Financial Trading System. CIDM 2007: 720-725

This paper presents a novel framework forpredicting stock trends and making financial trading decisionsbased on a combination of Data and Text Mining techniques.The prediction models of the proposed system are based on thetextual content of time-stamped web documents in addition totraditional numerical time series data, which is also availablefrom the Web. The financial trading system based on the modelpredictions (ADMIRAL) is using three different tradingstrategies. In this paper, the ADMIRAL system is simulatedand evaluated on real-world series of news stories and stocksdata using the C4.5 Decision Tree Induction Algorithm. Themain performance measures are the predictive accuracy of theinduced models and, more importantly, the profitability of eachtrading strategy using these predictions.

内容メモ†

1. その日のニュースをオンライン(reuter, forbes)から取得．
2. 決定木で24時間後の株価を予測．
   • オンラインのデータをストックする必要がない．その場で手に入るデータのみで予測可能！
     • とはいえ，決定木の構築には過去のデータが必要．