本研究結合由Holland(1975)所發展的遺傳程式與資訊科技的計算理論來研究臺灣股票市場的市場擇時策略，並根據Bauer(1994)對於投資法則的編碼方式，分別考慮以基本分析與技術分析為基礎的遺傳程式市場擇時策略，並且以買入持有策略、投資無風險資產與追漲殺跌策略等三種投資策略作為比較基礎來評估遺傳程式在投資策略應用上的可行性。 　　研究期間為1984年至1991年，其中樣本內期間為1984年至1988年，而樣本外期間為1989年至1991年，研究結論如下： 　　1.不論採基本分析或技術分析為基礎的遺傳程式投資法則，在樣本外期間第一年（1989年）的報酬均顯著高於投資無險資產與追漲殺跌策略，但低於買入持有策略。 　　2.就整個樣本外期間（1989-91年）而言，採用基本分析的遺傳程式投資法則顯著優於買入持有策略，而採用技術分析的遺傳程式投資法則並不具有投資績效。 　　3.以基本分析為基礎的遺傳程式投資法則較適用於長期投資，而以技術分析為基礎的遺傳程式投資法則較適用於短期投資。 　　4.樣本外期間經歷了遺傳程式沒有學習過的資料型態（亦即1990年初股價連續下跌趨勢），對於遺傳程式的學習能力形成了一大挑戰。

本論文是承續近來〝有限理性總體經濟學〞發展下之一支研究。有關有限理性的定義，在本研究中乃是以Sargent(1993)及Leijonhufvud(1993)為根據。Sargent(1993)認為：經濟學家在建立模型時，要怎麼樣去塑造其模型中的決策者的預期及學習呢？為了在精神上求一致起見，不應將模型中的決策者想成比經濟學家本人更聰明或更無知。有關這兩個角色應一致的要求，似乎便成了有限理性總體經濟學中相當關鍵的磐石。有關預期與學習形成的部份在計量經濟學上，又可大致分為兩個階段。在第一階段中，是以統計決策理論為主所建構的預期與學習過程，這類型的預期是奠基於以機率模型為主的學習過程。此類學習過程可以說是1980年代以來，理性預期學習過程的主要架構。使用這種學習模型需對決策者在所擁有的資訊上，做較強的限制。而第二階段的學習模式是要減輕模型中對決策人在資訊上的負荷，即將第一階段機率模型的學習擴充至非機率模型的學習。而幾乎所有學習上的問題，都可以視為一個尋找的問題，模型選擇是尋找模型，參數估計是尋找參數。在模型的設定上，以往我們處理的程序是：假設模型為....，則我們可以....。對於模型的選定並沒有嚴格的判定標準可供依循。然而遺傳規畫不但對模型的設立，提供了一個良好的典範，而且對如何尋找模型，提供了一個一般性的尋找模式。模型的選取，應是先經由尋找的過程而得到的，而非憑空自上帝的手中取得。因此，就如何建立起尋找的方式，其較模型的選擇更為基本且更為重要。遺傳規畫運作之初，並沒有包含先驗的知識，初始的模型是經由隨機創造而得。在演化的過程中，模型逐漸地有了系統（型態）的出現。這種尋找的過程，既不偏向隨機也不偏向系統，在隨機與系統中，取得了一個完美的平衡點。在遺傳規畫運作下，要選擇何種模型，將視實驗者的時間成本而定。換句話說，即遺傳規畫提供了實驗者到目前為止最好的模型，是否該花更多時間以取得〝較精確〞的模型，將由實驗者自行決定。在此情況下，我們在模型的選擇上，有了一個較為適當的判定基準：模型的大體輪廓將是藉由進化的方式取得，不是經由天外神來之筆而誕生。在模型精確度的選擇上，將由個人的時間成本來定奪。就在這層意義上來說，此種選擇的模式比較符合〝人性〞，亦與經濟學的精神相符合。本論文的目的便是要了解遺傳規畫在實際運作上的一些特性，以及該如何正確地使用它才能得到最大的功效，以期望它能成為我們在處理有限理性總體經濟學上的一個重要工具。

The existence of sunspots or sunspot equilibria has been debated for several decades on its influence in the field of Economics. While models of sunspots or sunspot equilibria have fitted well for some subsets of empirical features, it comes at a cost of moving further away from economic believability and robustness. Studies on the theoretical plausibility of sunspot equilibria have been addressed extensively in several different economic models, but exist almost entirely within the framework of the homogeneous rational expectations equilibrium devised of representative agents. This framework shapes later arising learning approaches to sunspot equilibria. These models have proposed various ways of learning, but they deal mainly with the learning of representative agents. Models of adaptive learning with heterogeneous agents, however, enable us to explicitly tackle coordination issues, such as the coordination mechanism of expectations. This is certainly desirable since sunspots are often used as a coordination device of expectations. In this dissertation, we continue this line of research, investigating the plausibility of sunspot equilibria in stock markets within the framework of heterogeneous agents and the dynamic relationship between sunspot variables and stock returns. We adopt an Agent-based Computational Approach, now known as Agent-based Computational Economics or ACE, to study the plausibility of sunspot equilibria. More specifically, we deal with this issue in the context of an Agent-based Artificial Stock Market (AASM). We contemplate AASMs to be highly suitable to the issue we examine here. Currently, none of the theoretical, empirical, experimental, or simulation models of sunspot equilibria directly capture sunspots within a stock market composed of heterogeneous agents. We conducted three series of experiments to examine this issue. From the results of these three series of simulations, we observed that sunspot variables generally do not have influence on market dynamics. This indicates that sunspot variables remain largely exogenous to the system. Furthermore, we traced the evolution of agents' beliefs and examined their consistency with the observed aggregate market behavior. Additionally, this dissertation takes the advantage of and investigates the micro-macro relationship within the market. We argue that a full understanding of the dynamic linkage between sunspot variables and stock returns cannot be accomplished unless the feedback relationship between individual behaviors, at the micro view, and aggregate phenomena, at the macro view, is well understood / The existence of sunspots or sunspot equilibria has been debated for several decades on its influence in the field of Economics. While models of sunspots or sunspot equilibria have fitted well for some subsets of empirical features, it comes at a cost of moving further away from economic believability and robustness. Studies on the theoretical plausibility of sunspot equilibria have been addressed extensively in several different economic models, but exist almost entirely within the framework of the homogeneous rational expectations equilibrium devised of representative agents. This framework shapes later arising learning approaches to sunspot equilibria. These models have proposed various ways of learning, but they deal mainly with the learning of representative agents. Models of adaptive learning with heterogeneous agents, however, enable us to explicitly tackle coordination issues, such as the coordination mechanism of expectations. This is certainly desirable since sunspots are often used as a coordination device of expectations. In this dissertation, we continue this line of research, investigating the plausibility of sunspot equilibria in stock markets within the framework of heterogeneous agents and the dynamic relationship between sunspot variables and stock returns. We adopt an Agent-based Computational Approach, now known as Agent-based Computational Economics or ACE, to study the plausibility of sunspot equilibria. More specifically, we deal with this issue in the context of an Agent-based Artificial Stock Market (AASM). We contemplate AASMs to be highly suitable to the issue we examine here. Currently, none of the theoretical, empirical, experimental, or simulation models of sunspot equilibria directly capture sunspots within a stock market composed of heterogeneous agents. We conducted three series of experiments to examine this issue. From the results of these three series of simulations, we observed that sunspot variables generally do not have influence on market dynamics. This indicates that sunspot variables remain largely exogenous to the system. Furthermore, we traced the evolution of agents' beliefs and examined their consistency with the observed aggregate market behavior. Additionally, this dissertation takes the advantage of and investigates the micro-macro relationship within the market. We argue that a full understanding of the dynamic linkage between sunspot variables and stock returns cannot be accomplished unless the feedback relationship between individual behaviors, at the micro view, and aggregate phenomena, at the macro view, is well understood.

為了瞭解影響演化性神經網路（ENT）預測表現的四項重要的機制：輸入資料性質、訓練樣本大小、網路搜尋密度以及控制模型複雜度，進而找出能使ENT充分發揮效果的組合。在本論文中首先設計ENT在模擬資料上的實驗，探討上述四項機制個別對預測表現的影響，再依照實驗結果的建議，設計能讓ENT發揮功效的組合，並以實際金融高頻率資料：香港恆生指數在一九九八年十二月報酬率為標的，探討模擬資料的結果在實際金融資料需要調整的部份。實驗結果顯示，當輸入資料經過線性過濾後，搭配大樣本訓練、高搜尋強度與適當地模型複雜度控制，會是能讓神經網路提高預測能力的組合。在實際金融資料的實驗當中同時發現，資料中偶而出現特別高或特別低的變化，會對ENT的預測表現有相當程度的影響。 / In this thesis, Evolutionary Neural Trees (ENTs) are applied to forecast the artificial data generated by financial and chaos models — iid random, linear process (Auto Regressive-Moving Average；ARMA), nonlinear processes (AutoRegressive Conditional Heteroskedasticity；ARCH, General AutoRegressive Conditional Heteroskedasticity；GARCH, Bilinear), mixed linear and nonlinear process (AR and GARCH). Experiments of the artificial data were conducted to understand the characteristics of ENTs mechanism. – data pre-processing procedures, search intensity, sample size and complexity regularization. From the experiment results of artificial data, the combination of pure linear or nonlinear time series, large sample size, intensive search and simple neural trees are suggested for the parameters setting of ENTs. And for the sake of computational burden, we have a trade-off between search intensity and sample size. Ten experiments are designed for ENTs modeling on the high-frequency stock returns of Heng Sheng stock index on December, 1998, in order to have an efficient combination of the factors of ENTs. The results show that ENTs would perform more efficiently if data are pre-processed by a linear filter, for ENTs will concentrate on searching in the space of nonlinear signals. Also, as is well demonstrated in this study, the infrequent bursts (outliers) appearing in the data set can be very disturbing for the ENTs modeling.