本研究以Hadoop平台為基礎讓股票先透過技術指標公式平行運算後,接著將K-means演算法套用於MapReduce框架上,藉此將股票作分群並同時提高運算效率,最後將分群結果定義決策後,再推薦投資者作買進或賣出之決策。本研究實證結果分群後的群集與大盤經過檢定後,大部分的檢定結果皆優於大盤,且能夠獲得更高之獲利,其分析出的結果呈現出來供投資者作為決策參考。
This research applies map-reduce parallel computing technologies to analyze the stock technical indicators on Hadoop platform. The computation efficiency is improved significantly; in the meantime, the target indicators are clustered by parallel K-means clustering algorithm and patterns are defined. Based on the found patterns, the most profitable buy-sell decisions will be recommended. The experiments were carried out to validate the proposed framework. Results show that most suggested buy-sell strategies beat the market and gain higher profit. In addition, the analyzed results could be used as decision support for stock investors.