本研究主要目標是以基因表達規劃法(Gene Expression Programming, GEP)多基因的設計，解決外匯保證金交易中，最難以解決的最適資金投入比率問題。本研究透過GEP多基因的設計，以及GEP隨機數值常數(RNC)權重設計機制，創新地以動態方式解決外匯保證金交易(Foreign exchange margin trading)資金投入比率問題，期望能避免人性因素而影響投資行為，並能穩定地獲得投資報酬率。經由研究結果顯示：(1)本研究所提出之技術指標模型，確實有機會為投資者帶來尚可的報酬，但在發生非預期性之經濟數據發佈時，則無法發揮該有的作用。(2)本研究採用的九項技術指標中，根據GEP之演化以KD與RSI較能為模型帶來超額報酬，而MA、MACD、Bollinger Bands等技術指標則較不具參考價值。

The main purpose of this study is to use the concept of collective intelligence to dynamically determine the optimal ratio of capital investment through the multiple-gene design of gene expression programming (GEP), and the weighting mechanism of GEP random numeric constants (RNC), hoping that the investment behavior to avoid been affected by human factors and obtain stable ROI..The results of studies show that: (1) The more stringent the trading signals are the worse it functions. 70-30 are more likely to bring stable excess returns for investors. (2) The model proposed in this study does have a good opportunity to bring acceptable rewards for investors. But it cannot function as sell as it is supposed in the event that unexpected economy data out of expectation are announced. (3) Among the nine technical indicators used in this study, KD and RSI are more likely to bring excess returns for the model according to the evolution of GEP, while MA, MACD, and Bollinger bands have less value for reference.