Ebru Efeoğlu

Terahertz imaging technology is a promising method for preventing terrorist attacks due to its ability to penetrate various materials and reveal hidden objects without emitting harmful radiation. This study proposes a fast classification framework with high detection accuracy for terahertz imaging systems that can be used to detect hidden weapons on people at airports and other checkpoints. The proposed framework is a hybrid of a deep neural network based on Color Layout Filter (CLF) and LSTM (Long Short-Term Memory). The performance of this neural network, combined with different optimization methods, was compared with classical machine learning techniques. In the performance evaluation, it was found that the deep neural network model using the Adam optimization algorithm outperformed classical machine learning techniques. This proposed neural network achieved 0.8890 accuracy, 0.8992precision, 0.8890recall, 0.8874F-score, and an ROC area of 0.9675after cross-validation.

One of the most visible symptoms of autism spectrum disorder is difficulty in speech and language. Difficulties in speech and language are generally very different for each child with autism spectrum disorder. Although some children with autism spectrum disorder can speak fluently, others will not be able to speak normally or will be even nonverbal. In all cases, parents try to communicate with, and understand their children’s needs, desires, and emotions. If a child with autism spectrum disorder cannot speak out loud, it is harder to communicate with him/her but there are other non-vocal methods for communication. In this paper, the benefits of teaching American sign language to children with autism spectrum disorder, the difficulties that families and children will experience while doing this, and technological solutions to these difficulties are presented. In parallel with advancements in technology, novel solutions to understand and use sign language have been proposed and these solutions are supposed to help parents who cannot understand sign language. Such solutions typically rely on image processing methods and classification algorithms to recognise sign language. Therefore, in this paper, the performance of various classification algorithms used to classify American Sign Language phonemes is compared. As the results show, when combined with image processing methods, classification algorithms can be used in various technological solutions aiming at helping to identify sign language phonemes.

Knowing the type of buried object before excavation prevents unnecessary excavation. Moreover, it saves time and money. In this study, an experiment set was prepared for the detection of buried objects. The experimental set was composed of an antenna that sends and receives electromagnetic waves in a wide frequency band, software that records and processes reflections, and a sandbox. In the study, metallic and non-metallic objects with different depths, sizes and shapes were buried in this sand pool and measurements were taken along a profile. 2D images were created from the measurements and image processing techniques were applied to these images. Classification algorithms were used to detect the type of bruied object from processed images. To increase the success of the algorithms, correlation-based attribute selection (CFS) and Principal Component Analysis (PCA) were used as attribute selection techniques. Genetic algorithm (GA), Particle Swarm Optimization (PSO), Harmony search (HA), and Evolutionary search (EA), which are among the metaheuristic optimization algorithms, were preferred as search methods in attribute selection with CFS. The performance of the algorithms was analyzed using the 10-fold cross-validation method. As a result, it was understood that the use of the PCA algorithm in attribute selection increases the classification success more than metaheuristic algorithms. The most successful among the classification algorithms used is the Random tree algorithm. After PCA, the accuracy value of this algorithm was 95.8 Therefore, a hybrid approach is proposed in which PCA and Random tree algorithms are used in the software embedded in the measurement system.

Disasters are events that affect life activities that cause physical, economic and social losses. These events cause loss of life and property, as well as damage to structures such as schools and hospitals that will affect the continuation of education and health services. There are two types of disasters. The first is man-made disasters and the second is natural disasters. Natural disasters occur as a result of natural events. Earthquake is a natural disaster. Disaster management is a process that covers pre-disaster, disaster and post-disaster. This study focuses on pre-earthquake disaster management. Safe construction is necessary to reduce the effects of earthquakes. Soil class is very important in a safe construction. Soil classification was made according to TBDY-2018 by using machine learning techniques for a safe construction in the Mediterranean region. 12 different machine learning algorithms were used for Classification and the results were analyzed. As a result of the analysis, the accuracy values of the algorithms are respectively: Naive Bayes 87%, LDA 88%, KNN 84%, Adaboost 96%, Logit boost 95%, Ultraboost 92%, BF Tree 98%, Extra Tree 84%, Random Forest 93%, Random Tree%. 95, Rep Tree 96%, SimpleCart 98%. The most successful algorithms in classification are Simle Cart and BT tree. The least successful algorithm is the Extra Tree algorithm.

Kaya sınıflandırması tünel ve yeraltı çalışmalarının yanı sıra madencilik ve hidrokarbon araştırmalarında da önem teşkil eder. Kayaların türünün yanlış tespiti hem para hem de zaman kaybına yol açar. Bu çalışmada 4 farklı kaya sınıfının fiziksel ve mekanik özellikleri kullanılarak kaya sınıflandırması yapılmıştır. Sınıflandırma için Bulanık Sırasız Kural İndüksiyon Algoritması (FURIA), Bulanık Kafes Akıl Yürütme (FLR), Çok Amaçlı Evrimsel Bulanık (MOE Fuzzy) sınıflandırıcıları kullanıldı. Çok Amaçlı Evrimsel Bulanık MOE Bulanık sınıflandırıcısı, ENORA ve NSGA II algoritmalarına dayanmaktadır. Bu nedenle bu algoritmalar ayrı ayrı kullanıldı. Verilere sınıflandırma öncesi Sentetik Azınlık Aşırı Örnekleme Tekniği uygulandı ve bu işlemden önceki ve işlem sonrasında sınıflandırma performansları karşılaştırıldı. Sonuç olarak Sentetik Azınlık Aşırı Örnekleme Tekniği ile sınıflandırma başarısının arttığı görülmüştür. Sınıflandırmada en başarılı algoritma FURIA algoritması oldu. Algoritma sınıflandırmayı %93 doğrulukla ve 0.16 hata değeri ile gerçekleştirdi.