本研究以國中一年級地理「台灣天氣與氣候」單元為例,探討貝氏網路的機率推論功能,在國中社會領域地理科是否能發揮預測及診斷學生錯誤類型、子技能及單元目標的功能,並以貝氏網路為基礎,配合順序理論發展出一套適用於此單元的電腦適性診斷測驗及電腦適性補救教學系統,並探討此系統的診斷精準度及補救教學成效。 研究結果發現: 一、 在利用貝氏網路診斷「台灣天氣與氣候」單元之錯誤類型、子技能及單元目標的精確度部分,與專家人工判定有高達97.23%吻合程度,顯示出優異的診斷效果。 二、 此套電腦適性診斷測驗系統約可節省三成左右的題目,且試題作答反應的平均 預測準確率達九成五以上,顯示該適性測驗的預測能力與完整測驗間的符合程 度相當高。在錯誤類型、子技能及單元目標診斷的平均預測準確率亦達九成五以上,表示適性測驗在錯誤類型、子技能及單元目標的判別與完整測驗有相當高的符合程度,亦即將貝氏網路推論理論應用於電腦適性診斷測驗的成效良好。 三、 此套電腦適性補救教學系統讓將近九成五的受試學生成績進步,大幅降低錯誤 類型發生及提升子技能達成。且所有受試學生的前、後測平均成績經T檢定後,達顯著差異。而針對高分組、中分組與低分組群等不同能力的受試學生做前、後測成績的T檢定後,亦達顯著差異。另外特別針對同分數但犯不同錯誤概念的學生,適性化地進行不同的補救教學,確實發揮因材施教的功效。
The study was to develop a computerized adaptive diagnostic testing and computerized adaptive remedial instruction system based on the probabilistic reasoning of Bayesian Network and ordering theory(OT) into Weather And Climate In Taiwan in first-grade geography course to induce the error types, sub-skills and unit target and to validate the effect of the system. The results of this study are as follow. First, Bayesian Network was employed to diagnose accuracy of the error types, sub-skills and unit target in Weather And Climate In Taiwan. The results had the 97.23% correspondence to the expert judgment, which show excellent diagnostic effects. Second, the computerized adaptive diagnostic testing system could save 30% of test items. The reaction of the test answers could be predicted and the accuracy achieved 95%, which showed the predictive ability of this system was similar to the complete test. The accuracy rate of prediction on the error types, sub-skills and unit purposes was over 95%, showing that the computerized adaptive diagnostic testing highly corresponded to the diagnosis in error types, sub-skills and unit target of the complete test. Apparently, the effect of the system was quite satisfying. Third, with the design, almost 95% of the students progressed in studies. The error types were decreased and the sub-skills were increased. All students’ average achievements on pre-tests and the post-tests had significant differences. The significant differences were also shown among the high-score, medium-score and low-score groups in the computerized adaptive diagnostic testing. The testers could receive the proper adaptive remedial instruction in time and enhance their learning.