本論文提出數個線性規劃模型建立成長基金的投資組合。目標函數皆以目標規劃方式呈現。第一個模型採用追蹤與成長差距最小的原則。第二個模型改採用大中取小原則。第三個模型則考慮時間因素對於投資組合的影響，修正第一個模型加入時間參數。最後以台灣上市股票市場作為實證分析對象，探討三組模型之表現。 / This thesis presents three linear programming models for selection of the growth portfolio based on historical data. The objective functions of these models are described by goal programming. The first model employs the principle of minimizing the deviation of the value-increasing index. The second model employs the mini-max principle. The third model is derived from the first model and includes the timing effect of historical data during construction of portfolio. The computational results and performance are illustrated by modeling with realistic data from the Taiwan stock market.

本論文使用大中取小原則及目標規劃技術，提出建構投資組合的數學規劃模型。要求此投資組合面對於不確定的股市，能夠在控制風險最小的情況下穩定且具有成長性的獲利。論文內探討如何透過數學的限制式來控制風險，而又能兼顧穩定且具有成長性的獲利，同時模型也可針對不同投資者的需求設定其數學規劃模型。最後以台灣股票市場做為實證分析的對象，給予不同的參數設定來驗證投資組合的表現。實證發現若以期初的配置比重持有到投資期間結束，此投資方式的績效欠佳。因此論文中進一步探討最佳的調整週期，實證顯示每經過8週，根據最新的資訊，重新調整建立新的投資組合，投資績效最好。 / This thesis proposed a mathematic programming model to construct a growth portfolio by using the mini-max principle and goal programming technique. The constructed portfolio is required to minimize the risk and to earn a stable profit under uncertain market. In the thesis, we discussed how to control the risk and maintain the growth of the portfolio by using the linear constraints. The proposed model also provides several parameters setting to meet the different investors' requirement. Finally, an empirical study will be provided by using the data from Taiwan’s stock market. The portfolios are constructed by giving different parameters and the performances are reported. The empirical study showed that holding a portfolio through the entire investment period without rebalance yield the performances that are not good. Therefore, the rebalance timing is investigated and the empirical study showed that a portfolio with rebalance strategy by every 8 weeks yield the best performance.

本文提出一個新的混合整數線性規劃模型建立投資組合。這個模型所採用的風險函數為最大損失的絕對值，而不是一般常用的損失變異數。在給定的報酬水準下，模型尋找在觀測期間中最小的最大損失的投資組合，即為大中取小的原則。模型也同時考慮實務上常遇見之情況，如：交易成本、最小交易單位、固定交易費用比率、資產總類數等限制。因此，模型內需使用整數變數及二元變數，導致模型的計算求解過程變得比不含整數變數及二元變數的模型困難許多。我們以固定整數變數的啟發式演算法增進求解的效率，並以台灣股票市場的資料做為實證計算的對象。 / A new mixed integer linear program (MILP) for selecting portfolio based on historical return is proposed. This model uses the downside risk rather than the variance as a risk measure. The portfolio is chosen that minimizes the maximum downside risk over all past observation periods to reach a given return level. That is a mini-max principle. The model incorporates the practical characteristics such as transaction costs, minimum transaction units, fixed proportional transaction rates, and cardinality constraint. For this reason a set of integer variables and binary variables are introduced. The introduction, however, increases the computational complexity in model solution. Due to the difficulty of the MILP problem, a heuristic algorithm has been developed for the solution. The computational results are presented by applying the model to the Taiwan stock market.