| dc.description.abstract | In the machine learning field used for gender classification through facial images, feature extraction is one of the inseparable parts. Several types of features can be extracted from images, such as texture features. In previous gender classification studies, LBG-VQ and MB-LBP can be used to extract texture features from images. When applied to the FEI facial images dataset, LBG-VQ produces suboptimal performance. Meanwhile, when applied to the FERET facial images dataset, MB-LBP produces optimal performance. Therefore, this study was conducted to discover the gender classification performance of the LBG-VQ and MB-LBP methods on the FEI facial images dataset. This study was carried out by implementing the two feature extraction methods separately or in combination with the two. The results of feature extraction were implemented into several classification methods, namely Naive Bayes, SVM, KNN, Random Forest, and Logistic Regression. The K-Fold Cross Validation was used to evaluate the model by splitting the data into 9 folds of data training and 1 fold of data testing. The combination of the Logistic Regression with MB-LBP 12x12 on LBG-VQ quantized images forms the machine learning model with the most optimal performance at 91,928%. To get the most optimal performance there are 3 preprocessing processes, namely noise removal, illumination adjustment, and image conversion to grayscale. Furthermore, there is also a second scenario of preprocessing that consists of first preprocessing processes with additional processes, namely face cropping and size normalization. The best optimal results with the second scenario of preprocessing is at 89,977% using SVM Poly Kernel and MB-LBP 12x12 on LBG-VQ quantized images. In conclusion, roundly the implementation of MB-LBP tends to show an increase in performance than the implementation of LBG-VQ. | en_US |
