Analisis Perbandingan Akurasi Pre-Trained Convolutional Neural Network Untuk Klasifikasi Kelompok Usia Pengunjung Rumah Sakit

Abstract

Children are not allowed to visit the hospital. Children should not visit the hospital for two reasons, namely the patient's side and the child's side. On the patient's side, patients need peace of mind during treatment and recovery. The noise generated by children makes the atmosphere not conducive and increases the patient's stress level. On the child's side, there are two factors, namely immunity, and trauma. Children have incomplete immunity so they are easily infected by viruses and bacteria. A child's immune disorder will harm the child's development. Apart from viruses and bacteria, in hospitals, there are also patients with major injuries such as those resulting from accidents. Children who see these large wounds can traumatize themselves and interfere with the child's growth and development. The age classification of visitors supports for hospital management to limit visitors based on age. Visitors categorized as children are visitors aged 12 years or younger. The method used for age group classification is the pre-trained CNN, including Alexnet, VGGNet, GoogleNet, ResNet, and AqueezeNet. We conducted a preliminary study using the All-Age-Faces (AAF) dataset as test data that represents the age of hospital visitors. The dataset is divided into two classes, namely children and adults. Based on the SqueezeNet test, it is a better method in terms of training accuracy and validation. Based on the order of accuracy validation, SqueezeNet succeeded in recognizing age groups with an accuracy of 93.09%, VGGNet 92.72%, AlexNet 91.44%, GoogleNet 90.92%, and ResNet 90.62%. This research is expected to contribute to helping control visitors to the hospital.