選擇就讀高級中學的學生首要目標就是考上理想的大學繼續升學,以現今大學多元入學方案來說,舉凡公立和私立高級中等學校的畢業學生或是具有同等學力證明的學生,通過資格審查後,才得以參加大考中心舉辦之「學科能力測驗和指定科目考試」,以及術科考試之各項成績參加招生,其中又以學測級分是多數學校採計衡量學生學習高中課程程度的重要參考指標。 本研究希望能從探討高中生的學測級分及其模擬考級分的關聯研究,做為日後輔導學生及招生相關事務之參考,蒐集高雄市某公立高中102年度、103年度及104年度畢業之高三應屆學生的大學學科能力測驗級分與在學模擬考級分,期望能由兩者的關聯性,準確的預測學測級分之落點,以提供學生及老師於推薦甄選時運用。 本研究主要使用倒傳遞類神經網路訓練於預測高中學生大學學科能力測驗級分,研究結論顯示:(1)運用模擬考總級分來預測學測總級分的準確度比運用模擬考各科級分來預測學測各科級分的準確度還要高,(2)運用模擬考各科級分來預測學測各科級分時,以社會科的準確度較高,(3)運用模擬考各科級分來預測學測各科級分時,以數學科的變異較大,結果也最差。
The primary goal of most students who have chosen attending the high school is to find an ideal university for their continuous studies after graduation. Students who are graduating from public and private senior high schools or those who have equivalent qualification can take either one or both the college entrance exams including "subject ability exam” and “designated course exam" organized by the college entrance exam center. The allocation of scores obtained from these two exams is usually hard to allocate. Therefore, building a forecasting model with good prediction is helpful for these students. The purpose of this study is to investigate the relationship between the scores obtained from three simulation tests of senior high school students and the scores obtained from college entrance exam of such students. By using the back propagation neural network forecasting, the objective of this study is to find out the suitable score allocations of the college entrance exam results for the senior high school students. In this study, a collection of three senior simulation tests results and corresponding subjective ability tests from a public high school in Kaohsiung for academic years 2013, 2014 and 2015 was investigated. The test results included the total scores for each simulation test and five indivisual scores for different courses such as Chinese literature, English, mathematics, nature science and social science. These tests results were designed to be the inputs of the neural network training. Another set of collections were the results for the subjevt ability exam from the same students. These tests results were desigeded to be the outputs of the neural network training. After training the back propagation neural networks, the best prediction of corresponding network was selected to be the prediction model for our research. The performances of both total scores of three simulation tests and individual scores of three simulation tests were investigated. From the results, we can conclude that (1) the performance of using total scores of the simulation tests results to predict the subject ability exam scores is much better than the performace of using individual scores of the simulation tests results, (2) the forecasting performance of using individual tests scores for social science is the most accuracte among the five courses, (3) the forecasting performance of using individual tests scores for mathematics is worse than other courses.