本研究旨在透過類神經網路中之多層函數連結網路(Multilayer Functional-Link Network,MFLN)以建構學生輔導類神經網路模式,並進行學生偏差行為的判斷與懲戒預測分析,依據預測結果,探討學生偏差行為與成因以及學生懲戒方式,據以提供教育相關單位、輔導人員及教師,作為預防性之學生輔導工作參考。 本研究以高雄市某私立高級職業學校90學年度至93學年度入學於軍訓室及輔導室之輔導學生數200人為研究對象,分為訓練組120人及測試組80人進行研究。學生輔導類神經網路模式之輸入層及輸出層分別為科類別、父母婚姻狀況、學生宗教信仰狀況、家庭經濟狀況、吸煙、打架、遲到、作弊、警告、記過、留校察看、輔導轉學、休學、個人因素、家庭因素、學校因素、社會因素等17項。本研究依各類神經網路模式所預測的結果歸納成四點: 一、學生偏差行為、偏差行為成因、懲戒方式會因為所就讀科類別之不同而有所不同。 二、學生偏差行為、偏差行為成因、懲戒方式會因為父母婚姻狀況之不同而有所不同。 三、學生偏差行為、偏差行為成因、懲戒方式會因為學生宗教信仰之 不同而有所不同。 四、學生偏差行為、偏差行為成因、懲戒方式會因為家庭經濟狀況之不同而有所不同。 根據預測分析結果,本研究之學生輔導類神經網路模式預測推論有八成以上的預測結果準確度,證明本研究之學生輔導類神經網路模式之可行性與實用性。
The purpose of this research is to build an artificial neural network model of students’ counseling, and we apply it to process prediction and analysis of the estimates of students’ deviant behavior and the punishment for their deviant behavior. According to the result of the prediction, we discuss the causes of students’ deviant behavior and the ways of punishment for their deviant behavior. We could also provide the data for institutions, counselors or teachers as reference materials for counseling. In this study, there were 200 students counseled by the counselors in Military training office and Counselors’ office at one of private vocational senior high schools in Kaohsiung City during the school year of 90 to 93. We divided them into two groups, 120 people in the training group and 80 people in the testing group. The input and output layer consists of 17 inputs which are departments, marriage status of parents, economics, religions, smoking, fighting, being late for school, cheating on tests, counseling, demerits, academic probation, transferring to another school, quitting school, four factors of individuals, families, schools, and societies. Based on our reservation on the outcome of the prediction of artificial network models of the research, we could generalize four causes which are departments students attended, marriage status of parents, religions of students, and home economics conditions. As far as the deviant behavior of students, the causes of deviant behavior of students, and the ways of punishment for the deviant behavior of students are concerned, they vary according to these four causes. The result of the experiment was that we obtained over 80 percent of the accuracy of the prediction, and it shows the feasibility and the practicality of this model.