查詢結果分析
來源資料
頁籤選單縮合
題 名 | 以電腦模擬探討群体決策模式在設計評價的應用=A Study on the Application of Group-Decision Techniques to Design Evaluation Through Computer Simulations |
---|---|
作 者 | 蔡登傳; | 書刊名 | 科技學刊 |
卷 期 | 7:1 1998.01[民87.01] |
頁 次 | 頁53-66 |
分類號 | 440.8 |
關鍵詞 | 設計評價; 產品評價; 群体決策; Design evaluation; Product evaluation; Group decision; |
語 文 | 中文(Chinese) |
中文摘要 | 群體決策是一種集多數人的個別意見而作成群體共同意見的決策技法。這種技法很適合應用在設計評價上。以群體決策技法評價的基本程序為:由評選者從受評產品中選出入選的優良產品若干件,依此即可綜合評選結果,排定各產品的名次。本文將此種程序稱為單向式評選法。上述方法中,若評選者也同時選出不良的產品,則將之稱為雙向式評選法。本文即探討這兩種不同評選法中評選人數、受評產品數、及每人可選出的產品數等三項因素的變化對評價結果的影響。首先,假設一組已依等第(即所謂的好壞)排序的產品,各產品的排序各不相同。因而排序在前的產品被選為優良的機率高於排序在後的,反之,被選為不良的機率低於排序在後的,準此原則訂定各產品的被選機率,以電腦進行評選模擬。模擬時,評選人數由1至50人,受評產品件數由5至50件,選出件數在單向評選法時為1至總產品件數少1件,在雙向評法時為1至總產品件數的一半。將模擬結果的排序與上述假設的產品等第排序作相關分析,其相關係數可解釋為該評選的精確值。各次模擬的精確值經分析後,可歸納出下列數項結論:1)參與評選的人數愈多,評選結果愈精確。一般情形下,評選人數在8人以上時,其評選結果的精確值可達0.8以上。2)每人選出件數與評選精確值的關係,單向式與雙向式不同,單向式呈二次曲線關係,而雙向式則呈對數曲線增加。3)單向評選中,評選人數與評選精確值的最高值呈對數關係。以上所得各迴歸方程式可提供實際設計評價應用時的參考。 |
英文摘要 | A group-decision technique gathers individual opinions and makes a collect decision of the group. This technique is considered suitable for application in design evaluation. In group-decision evaluation, the evaluators are asked to select among a group of products the best designed ones. The results are then used to rank order the products according to their frequencies of being selected. This process of selecting good products is called a single-ended technique. If the evaluators have to select bad products as well as good ones, then it is called a double-ended technique. This paper investigates the effect of three variables in the evaluation process: m, the number of evaluators, n, the number of products to be evaluated, and t, the number of products to be selected as good (or bad) by the evaluators. At first, it is assumed that there is a set of products distinctively rank ordered according to their goodness in design. Hence, if product A is ranked prior to product B, then the probability of A being selected as good must be higher than that of B. Conversely, the probability of A being selected as bad must be lower than that of B. All products ucts are accordingly assigned their respective probabilities of being selected as good and bad for use in the computer simulation. In the simulation, the number of evaluators m ranges from 1 to 50, the number of products for evaluation n ranges from 5 to 50, and the number of products to be selected t ranges from 1 to n-1 in the case of single-ended evaluation, and from 1 to n/2 in the case of double-ended evaluation. The correlation coefficient of the result of a simulated evaluation and the hypothesized rank order can be considered as the precision of this evaluation. Analysis of the precision values reveals several phenomena as summarized in the following: 1) The more the number of evaluators, the higher the precision of the evaluation. In general, when the number of evaluators is equal to or greater than 8, a precision value of 0.8 or above can be obtained. 2) The relationship between the precision values of the evaluations and the numbers of selected products are of a quadratic curve in the case of single-ended evaluation and of a logarithmic curve in the case of double-ended evaluation. 3) In the case of single-ended evaluation, the relationship of the number of evaluators and the maximum values of precision values of evaluations are of a logarithmic curve. The above mentioned regression curves can be applied for design evaluation in real situations. |
本系統中英文摘要資訊取自各篇刊載內容。