非線性演化式股票評價與資金配置模型

Abstract

近年來台灣股票市場蓬勃發展，市值成長快速，報酬波動度(volatility)擴大，股票市場投資需要承擔相對較高的風險。股票評價的目的是評估股票的合理價格並挑選被市場低估的股票，搭配適當的資金配置以降低投資風險並獲取超額報酬。然而，兼具股票評價與資金分配的模型的研究並不多見。傳統評價方法存在許多模型的假設前題，不一定能有效符合市場特性；應用不同傳統模型於相同的證券市場也可能產生不同的研究結果，這表示股票的合理價格應該是區間值而不一定是單一值。傳統資金配置常見的作法是均等分配或依特定順序組合的線性分配模型，不適合應用於複雜、雜訊多且資訊多變的股票市場。本研究的目的是應用遺傳演算法建構「演化式股票評價」新模型，以最佳化股票價值的合理區間與交易區間。同時結合三次方程內插法，建置最適的非線性資金配置策略，以提高投資效益。研究結果發現，內含交易成本的變動型交易成立點（情境4）投資績效最佳，顯示本研究模型會依股票特性，演化出最適的合區股價區間與交易區間的資金配置策略，以提昇最終投資效益。此外，經由不同的盤勢分析發現，當大盤處於空頭時，股票愈低資金持有遞增分散持股成本；盤整走勢時，當股落入低估區與高估區才有交易行爲，無形減少交易次數與成本，此兩種走勢最能表現其獲利能力。整體而言，演化式股票評價的投資績效仍優於大盤。

Taiwan stock market has experienced higher volatility including stock prices and liquidity recently. It means that the stock investment will take higher risk with higher market volatility. Stock valuation is to calculate the required rate of return that company we are valuing. However, using different traditional valuation models will get unique risk/return combinations, even those in the same stock market or same industry. It suggests that the fair value of a stock should be a range rather than a single value. In this article, we propose a nonlinear stock investment and capital allocation model using a hybrid of Genetic algorithm (GA) and cubin spline (CS), where GA is used to optimize the fair range of stock price, and CS is used to estimate nonlinear capital allocations. From our experiments, this hybrid of GA and CS model could get better investment returns than using buy&hold generally. In addition, the mean stock price almost falls into the fair stock price range predicted by our model.