การเปรียบเทียบประสิทธิภาพระหว่างเทคนิค Decision Tree C45 k-NN และ Naive Bayes เพื่อใช้พยากรณ์ผลสอบมาตรฐานด้านเทคโนโลยีสารสนเทศ มหาวิทยาลัยราชภัฏนครปฐม
The performance comparison between Decision Tree C4.5, k-NN and Naive Bayes to predict the exit exam of information technology of Nakhon Pathom Rajabhat University.
Abstract:
This thesis aims to measure an effectiveness of data mining in forecasting a result of IT Standards Test for students of Nakhon Pathom Rajabhat University. Three data classification techniques consisting of Decision Tree C4.5, k-NN, and Naïve Bayes focus on this study by adopting RapidMiner Studio program as a tool in evaluating data. The population to be tested in this study is taken from the student database provided by the universitys register office from year 2009 to 2014, including student type, year, gender, age, faculty, education level, grade point average, and IT standards test score. Besides, 10-Fold Cross Validation is applied in an evaluation of model effectiveness for the purpose that the most accurate value could be achieved. The result of the study shows that Native Bayes technique delivers the best solution amongst all techniques with an accuracy level at 94.35% and absolute error at 0.059. Therefore, it could be concluded that Native Bayes is the most appropriate data classification techniques for Nakhon Pathom Rajabhat University