|
An Efficiency Comparison in Prediction of Outliers 6 Classifications |
|---|---|
| รหัสดีโอไอ | |
| Creator | Saichon Sinsomboonthong |
| Title | An Efficiency Comparison in Prediction of Outliers 6 Classifications |
| Publisher | Thammasat University |
| Publication Year | 2563 |
| Journal Title | Thai Journal of Science and Technology |
| Journal Vol. | 9 |
| Journal No. | 3 |
| Page no. | 255-268 |
| Keyword | k-nearest neighbor, artificial neural network, rule-based, binary logistic regression, voted perceptron, stochastic gradient descent |
| URL Website | https://www.tci-thaijo.org/ |
| Website title | THAIJO |
| ISSN | 2286-7333 |
| Abstract | In this study, an efficiency comparison in prediction of outliers 6 classifications were determined. The classification methods were compared the followings: (1) k-nearest neighbor method, (2) artificial neural network method, (3) rule-based method, (4) binary logistic regression method, (5) voted perceptron method, and (6) stochastic gradient descent method. The purposes were to compare the efficiency of 6 classifications, and to compare SPSS, MINITAB and WEKA programs. The following efficiency comparison values were employed, i.e. accuracy, mean square error (MSE), and mean absolute error (MAE). For the low outliers data set (0-3 percentage), banknote authentication, the best classification method was the stochastic gradient descent method in combination with the WEKA sampling method. The middle outliers data set (3-6 percentage), Facebook metrics, the best classification method was the k nearest neighbor method in combination with the WEKA sampling method. For the high outliers data set (6-10 percentage), contraceptive method choice, the best classification method was the artificial neural network method in combination with the WEKA sampling method. |
