|
|
 |
| |
|
|
|
Quality Control and Classification of Steel Plates Faults Using Data Mining |
|
|
|
PP: 59-67 |
|
|
|
doi:10.18576/amisl/060202
|
|
|
|
Author(s) |
|
|
|
Mohammad Ali Afshar Kazemi,
Sima Hajian,
Neda Kiani,
|
|
|
|
Abstract |
|
|
| Evaluating the quality of steel Plate is vital for factories and doing this manually is associated with life-threatening risks
and low efficiency. Given that the steel Plate faults often create problems in the production process and sometimes may lose market,
need to check the quality of the Plate surface and find the failures are inevitable. This paper tends to first identify and recognize the data
based on the CRISP-DM cycle in data mining and thus compares and evaluates four data mining models to classify seven commonly
occurring faults of the steel plate. For this purpose, the models of C5.0 decision tree, Multi Perception Neural Network (MLPNN),
Bayesian network (BN) and Ensemble model are used. A faults dataset of steel plates is taken from the University of California at
Irvine (UCI) machine learning repository. The diagnostic accuracy of C5.0 decision tree obtained remarkable performance with an
accuracy of 95.56 percent for the training data and the accuracy of 95.66 percent for testing data. The results of the model fitting of
training and testing data indicated that the model C5.0 is superior to Multi Perception Neural Network (MLPNN), Bayesian network
(BN) and Ensemble model. |
|
|
|
|
|
|
|
