本研究以展望理論中之價值函數概念修改TOPSIS （Technique for Order Preference by Similarity to Ideal Solution）模型，進而發展出具風險行為考量之兩類似距離為基礎之多階段多準則決策模型，並以台灣不同期間之全球型基金為研究案例，進行基金績效評估。 本研究利用TOPSIS中之距離函數概念，更換為展望理論中之S型價值函數，並以正反理想解PIS及NIS為參考點，以發展出具風險偏好之TOPSIS模型。接續以考量利得減去損失之淨利得概念，建構另一模型。為了解基金之長期績效，本研究實測2015年及2008年金融海嘯期間之全球基金資料，並運用打擊率及夏普指標，了解不同模型的績效表現。再使用統計檢定不同模型間之相關性，以及對權重進行敏感度分析與模擬分析，比較三種模型 (TOPSIS模型、具風險偏好之TOPSIS模型、距離基礎的淨利得模型) 之績效差異。 分析結果顯示，兩修改模型於月投資時在波動大的金融海嘯時期，表現明顯優於TOPSIS模型；而TOPSIS模型在月投資時，則適合波動小的一般時期。相較於月投資之基金績效表現，其季投資所得出之結果與月投資相反，且勝率較月投資高，為應應一般時期更有效的基金績效評比，三種模型更適用於季投資方式，其中兩修改模型表現優於TOPSIS模型。經統計檢定得知，兩修改模型之排序結果與TOPSIS模型相似，表示此三者具相同特性。另在敏感度分析與模擬分析中可看出，兩修改模型對權重的變化相當穩定。經以上分析可知，本研究所建構之模型具實務價值，適用於變動劇烈的金融市場上，能提供投資者決策之依據。 This study aims to improve the Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) model which is based on S-sharped value function in prospect theory, in order to establish two Distance-based Multi-stage MCDM Models included risk consideration. Examples of Global Fund are evaluated by two types of periods purchased in Taiwan. The study uses the concept of Distance-based in TOPSIS, transferring the S-sharped value function in prospect theory. To consider risk preference in TOPSIS, we employ PIS (positive ideal solution) and NIS (negative ideal solution) as two reference for gains and losses in the modified TOPSIS. Furthermore, we build another model which utilize the concept of net gain. In the direction of realizing the long-term fund performance, we take research data as Global Fund in the years of 2015 normal time and 2008 financial crisis. Then we use batting rate and Sharpe index to test fund performance in different periods, and in these three models (TOPSIS model, distance-based risk TOPSIS model, distance-based net gain model). We test correlation of three models, and compare them by adjusting weight criteria, such as sensitivity analysis and simulation. As a result the two modified models are more adequate in 2008 financial crisis case than TOPSIS model; however, TOPSIS model is fitted in 2015 normal time by month investment. Although the result in season investment is contrary to month investment, the performance in season investment is better than month investment. For effective investing funds in normal time, season investment is suitable for three methods to evaluate funds performances; especially, two modified models are better than TOPSIS model. Spearman test shows that the ranking of two modified models are correlated with TOPSIS model. Sensitivity analysis and simulation specify that two modified models are more stable when weight criteria changed. Through the study, the two modified models are regarded as practical tool and can be used in today’s financial market. It is effective for decision making to deal with fluctuation data.