本文以屬於人工智慧之一的基因演算法模擬臺灣發行量加權股價指數，並與相同權數及隨機權數之天真模擬及傳統二次規劃法模擬的指數作比較。結果發現，以簡單方式選定一組大樣本的股票環境中，基因演算法可模擬出與加權股價指數報酬率平均相關係數高達95% 以上的指數，優於其他所使用的方法，而篩選出的平均家數也較其他方法為少。在一定的前提假設下，模擬指數組合的風險也與市場風險相當。因此即使在大樣本環境中，實務界使用基因演算法來改進臺灣發行量加權股價指數的模擬應是可行的。

　　　　　This article experiments with genetic algorithms, a branch of artificial intelligence, in simulating Taiwan's weighted stock index, and compares the results with those by naive and quadratic programming simulation. Results indicate that with a simple method in stock sampling, the simulated index is highly correlated with the Taiwan's stock index in terms of returns. The overall correlation is over 95% and is better than other method used for comparison. Also, under certain presumptions, the simulated stock index is well diversified and has only the market risk. It is therefore feasible of using genetic algorithms in simulating stock index. In the future, if we modify the objective function to include the autocorrelation behavior of stock index, genetic algorithms should be an effective tool in simulating the Taiwan's stock index.