本研究首先建構四種擇股決策模型包括GP模型、GA模型、GD模型、NNW模型並探討各模型在投資組合決策績效表現。並將該四個模型所得的結果，結合級比檢驗並應用證據理論整合爲DS模型。本實驗設計針對台灣證劵交易所中電子工業類的上市公司進行擇股決策。由於本研究以基本面分析爲主，因此所使用的評估變數屬於財務比率，將原始資料經處理除去單位的影響後將研究樣本透過五種模型進行運算，求得最佳投資組合。 研究結果顯示各模型之平均報酬率經比較後，得知所有模型之平均數均高於台灣加權股價指數、電子類股價指數甚至高於五大銀行之三個月存款利率，表示無論選擇任何模型，均能帶給投資人的投資報酬遠高於定期存款利率提高財富價值。此外雖然GP模型灰色預測之績效未能超越全體樣本平均報酬率，但實證結果發現GP模型經級比檢驗後並藉由與其他模型之實證結果證據融合所建立之DS整合模型，報酬績效均較原四種基本模型之投資組合大幅提升，其中DS整合模型之總平均報酬更高達112.45%，由此可知灰色預測模型中利用級比檢驗降低誤差並對不同模型進行互補結合，對報酬率之提升確實有極大的幫助，同時所建構的擇股模型可發揮最大之預測綜效。

This research work applies four different approaches including Grey Prediction (GP), Neural Network (NNW), Genetic Algorithm (GA) and Grey Decision (GD) to construct sophisticated stock selecting models which are capable of providing profitable outcomes for investment and reducing the risks from human subjective judgment. Based on using financial ratios as criteria to evaluate the electronic industry companies listed in Taiwan Stock Exchange (TSEC) for portfolio selection, all designed models perform well on selecting promising stocks and the operational average returns exceed the defined benchmarks. The outcomes of four models are then combined and fused with Dempster Shafer combination rules and integrated as the named DS model. The total average return acquired from the DS model increases significantly to 112.45% and surpassed other models. The results show that the data fusion of four models is able to produce optimal synergy and demonstrates higher evaluation performance.