Arab Face Recognition and Identification Based on Ethnicity and Gender Using Machine Learning

Abstract

Researchers are highly interested in the classification of ethnicity using the human face since every individual has features that distinguish him from others, and every group of people shares some features that set them apart. These features are called ethnicity. A shortage of academic inquiry into the Arab world is well acknowledged. To achieve this, this research seeks to generate an Arab dataset by first grouping all Arab countries into similar categories and then classifying these labels using machine learning methods. The Arab face dataset created consists of five labels: Arab Gulf States, Egypt, Levant, Maghreb, and North and East Arab African Countries. This paper uses six types of Machine Learning to classify gender and ethnicity: Artificial Neural Network (ANN), logistic regression, Support Vector Machine (SVM), naïve bayes, K-Nearest Neighbors (KNNs), and random forest. SVM model has recorded the best result to classify gender and ethnicity with 92.7% Area Under the Curve (AUC) and 57.6% accuracy, and ANN model has recorded the best result to classify ethnicity with 92.2% AUC and 72.2% accuracy.