由於股市投資具備高報酬率的特性,但一般投資人在面對股市行情與資訊時無法合理的分析和判斷。本研究以半導體產業為研究對象,並以資料探勘的技術尋找股票市場中真正具有投資價值,利用亂數基礎分類法與粒子群優化算法建立模型後 並以大盤法和淨投資報酬率法將資料之準確性進行分析,建立有效的投資組合。整個研究設計包含兩部 : 以亂數基礎分類法先進行財務屬性的分析後,並以粒子群優化算法分析股票市場,以進行投資組合的選取。為了符合上述的要求,我們以進行四組分析: (a)原始財務屬性+大盤法 (b)原始財務屬性+淨投資報酬率法(c)屬性刪減+大盤法 (d)屬性刪減+淨投資報酬率法,將資料依據屬性分為訓練資料與測試資料,將各檔股票財務屬性正規化後,利用粒子群優化算法出建立選取投資組合的模型,並分析了模型的準確性。
Since the stock market has the characteristics of a high rate of return, the investors generally have trouble on analyzing the stock market information. In this study, we focused on semi-conductor stock market. The data mining techniques are adopted to find investment value of the stock market. The particle swarm optimization (PSO+kmeans) with entropy-based classification(EBC) is used to develop the stock market portfolio model. The Weighted Price Index of the Taiwan Stock and Return of Interest (ROI) are collected to investigate the analyzed model. The entire study has two stages. The EBC is used as feature extraction on finance attributes from stock company. Then, PSO+kmeans is used as clustering method to select stock company as target portfolio. There are four case studies: (a) Original data with Weighted Price Index of the Taiwan Stock (b) Original data with Return of Interest (c) feature extraction with Weighted Price Index of the Taiwan Stock (d) feature extraction with Return of Interest. All the finicial data is normalized and PSO+kmeans is used as selection tool for portfolio modelling . The model is evaluated and results of accuracy is investigated.