在情境感知無所不在學習環境中使用基因演算法優化學習活動的次序

在情境感知無所不在學習環境中，若能藉由個別化的導引和數位學習系統的輔助，依活動彼此間的關聯度，適當地安排學生去觀察體驗一序列有意義的學習活動，將可能有效的提升學習效果。在本研究中，我們提出了二個有關在情境感知無所不在學習環境中的最佳化學習次序問題，運用基因演算法的求解特性，希望找出基因演算法中較佳天擇、交配方式的組合，將之設計運用在在情境感知無所不在的學習活動中。為了要處理本研究中的最佳化問題，我們提出了二個目標函數，運用基因演算法中不同的天擇、交配方式，設計六種方法來加以驗證。此外，在問題一中我們另外設計 Heuristic 演算法和Random 演算法與基因演算法做比較，驗證何者的效能為佳。最後所得的實驗數據顯示，二個學習活動次序最佳化問題，在使用基因演算法演化運算的過程中，天擇階段的挑選機制使用競賽法（Tournament），交配階段使用循環交配法（CX）的組合，輔以交換突變的機制，是優化學習活動次序的有效方法。

關鍵字

無所不在學習；基因演算法；情境式學習；情境感知

並列摘要

In context-aware ubiquitous learning environments, with the help of individualized guidance and digital learning system, students can be arranged properly to observe and experience a series of meaningful learning activities according to the relevance among them. As such, the learning efficacy is enhanced. In this study, we propose two formulations for optimizing the sequence of the learning activities for each individual in context-aware ubiquitous learning environments. Based on the characteristics of genetic algorithm, we seek for the best combination of alternative implementations of Selection and Crossover in context-aware ubiquitous learning environments. In order to cope with this optimization problem, two objective functions are proposed and six combinations of algorithmic implementations are tested. In addition, the performance of our genetic algorithm is also compared with those of Heuristic and Random algorithms. According to the experimental result, the combination of Tournament Selection, Cycle Crossover, and Two-swap Mutation is the most effective strategy for optimizing the sequence of learning activities.