Abstrak – The recognition or classification of patterns is a major problem in computer vision. Many methods have been applied such as: moment invariant, Artificial Neural Networks (ANN), K-mean, Support Vector Machine (SVM) and others. These methods have a few limitations. The moment invariant fashion is highly vulnerable to noise. ANN methods require a long computing time (especially multi-layer ANN) during the training process. On the other hand, the dimensions of the features generated from the methods are relatively high, which requires large storage space (memory). In addition, this leads to the long computing time when the testing process is carried out. Based on these facts, this research makes use of methods that being able to reduce the feature dimensions, namely the Principal Component Analysis (PCA). In the PCA method the dimensions of the sample image are converted to principal components (face space), whose dimensions are much smaller than the dimensions of the sample image itself. Our works exhibit that the PCA method is highly effective in carrying out the pattern classification process. This can be indicated by the relatively high values of Predictive Accuracy, Precision and Recall (close to 1) while the FP Rate is low (close to 0). Moreover, the location of the point coordinates (FP Rate, TP Rate) in ROC graphs is fallen in the upper left region (approaching the perfect classifier region).
Copyrights © 2019