為達成組織的目標,組織必須經常進行適當的分組(Grouping),如何有效分組,乃成為一個重要的課題。儘管分組的問題一直受到高度的重視,但隨著社交網路(Social Networks)的蓬勃發展,人與人的互動方式隨之不同,社交關係也跟著改變,因此,必須重新思考其所帶來的影響。如何在考慮社交關係的影響下,建立一個最佳的分組機制,是當前非常迫切需要探討的議題。 分組最佳化必須深入考慮若干問題。首先,在進行分組的同時,應該要能發現許多隱藏在問題背後的資訊,例如,社交關係與分組結果的關聯性、選擇分組夥伴的主要理由等。其次,分組可分為同質的(Homogeneous)與異質的(Heterogeneous),一般而言,異質分組是個較佳的方式,可改善學習成效或社交關係。根據過去的研究可知,人際關係是最困擾學生的主要問題之一,如果能根據社交關係進行異質分組,可能有助於人際關係的改善。再者,學習成效(成績)也是學生最在意的問題之一,若能運用一個有效的方式進行分組而改善學習成效,將對教育有很大的幫助。此外,好的分組必須滿足其他的要求,例如每組必須有不同職位、或不同資歷…等等各種成員。有鑑於此,一個好的分組機制應能滿足這些組織特定的需求。 針對上述課題,本研究提出社交測量法與基因演算法(Sociometry and Genetic Algorithm, SGA)來進行分組最佳化,此方法藉由簡單的問卷來蒐集組織成員對其他成員的偏好(Preference)或志願(Choice),經彙整後運用社交測量法找出關係結構(Relationship Structure)及社交指數(Index of Sociometric Status Score, ISSS)並依據其決定出權重值,再運用基因演算法(Genetic Algorithm, GA)進行分組最佳化。 本研究透過分枝界限法(Branch and Bound, B&B)與基因演算法比較,結果發現基因演算法能得到與分枝界限法相同的結果,顯示分組系統的運算結果具有正確性。此外,本研究以三個真實的案例來進行分組實驗,分別是籃球團隊分組、國標舞分組及校外教學分組。實驗結果顯示,SGA確實能根據不同的組織要求有效地分組。再者,本研究經與隨機法比較分組滿意度及學習成效後,結果顯示採用SGA的滿意度顯著優於隨機分組法,且透過該方法能有效提升學生的學習成效,因此驗證本研究提出的方法確實有效。
In order to achieve the organization's goals, the organization must be grouping properly, how to grouping effectively is become the most important issue. Although grouping has been important issue, but with the social network raised successfully, the interpersonal interaction along with social relationships are changing, therefore, its rethink the future affect. How to establish a grouping optimization on considering social relationships that is currently urgent to investigate the issue. To attain the objective, the organization often assigns their individuals to different groups. Consequently, a mechanism that can best assign individuals to groups is greatly needed. Previous studies, we are proposes sociometry and genetic algorithm approach (SGA) to solve this kind of problem. First of all, members are asked to show their preferences to several preferred candidate groupmates. Then members’ preferences are aggregated and the sociometry is employed to find the relationship structure and to calculate index of sociometric status score (ISSS). The ISSS are calculated to as the weights of members according to different objectives and the genetic algorithm is employed to optimize the grouping. Grouping optimization is must to consider some problem. First , grouping must to discover lots of information which hide, for example, social relationships and the relevance of grouping results, the main reason of select partners. Secondly, grouping has two ways, including homogeneous and heterogeneous, and heterogeneous grouping is a better way to improve learning performance or social relationships. To validate the correctness of the GA program, branch-and-bound is also employed, and its results are compared with those from GA. Results from some experiments show that GA can obtain the same optimum solutions as those from branch-and-bound. In addition, to evaluate the effectiveness of SGA, three real cases were utilized, including basketball team grouping, ballroom dance grouping and outdoor education grouping. Experimental results show SGA can optimizes the grouping effectively according to organizational requirements. Besides, the SGA satisfaction and learning performance is significantly higher than the random method. Therefore, it is to verify we propose the SGA method with the effectiveness.