The Public Safety Network (PSN) is important for providing communication after a possible disaster. It provides a communication opportunity between the victims under the wreckage and the rescue team in the case of base stations do not function. Device-to-device (D2D) discovery and communication provides communication services in such difficult situations without the requirement of network infrastructure and by reducing call traffic in cellular communication networks. Therefore, device discovery according to different parameter values in the simulator environment for 3GPP and PSN provides great convenience to users in the application phase. In this study, we presented a device-to-device (D2D) discovery simulator based on the Matlab Graphical User Interface (GUI) for the students and researchers related to the wireless communication area. In this simulator, there are different pathloss models for D2D communication along with the parameters suggested by 3GPP and PSN for educational and research purposes. D2D discovery results are given in visual graphics with the parameter settings in the simulator’s interface. Thanks to the D2D discovery simulator, the students and researchers can easily understand how the user mobile devices discover each other which is necessary for D2D communication in the disaster area. They can also compare device discovery algorithms with different parameters and models. Six different scenario examples and an evaluation questionnaire about the D2D discovery simulator were proposed to students and researchers at a University in Turkey. The users (students and researchers) can easily test D2D discovery algorithms with different simulation parameters with a user-friendly interface thanks to the developed simulator.

Machine learning (ML) and deep learning (DL) have significantly advanced, enabling transformative applications in natural language processing, computer vision, and speech recognition. This progress demands accurate and efficient modeling, particularly for large datasets. In credit card fraud detection, diverse DL and ML approaches have been employed to enhance performance and reliability. This study utilized the Creditcard.csv dataset, consisting of 284807 European transactions with 30 PCA-transformed features and a ‘Class’ variable for legitimacy or fraud. The dataset was preprocessed and evaluated under both imbalanced and balanced conditions. Ten models, including Logistic Regression (LR), Naive Bayes (NB), K-Nearest Neighbors (KNN), Decision Trees (DT), Random Forest (RF), Support Vector Machine (SVM), XGBoost, Gradient Boosting (GB), LightGBM (LGBM), and Autoencoders (AE), were assessed using precision, recall, F1 score, and AUC. All experiments were conducted in Python. The findings demonstrate the strong fraud detection capabilities of LGBM (RandomOverSampler), RF, and KNN, while XGBoost (RandomOverSampler) achieved nearly perfect AUC. Ensemble learning methods such as RF, XGBoost, and LGBM demonstrated superior generalization and robustness. These methods consistently achieved accuracy scores above 0.99 and AUC values above 0.96, highlighting their effectiveness in addressing class imbalance and identifying complex fraud patterns.

The IoT-driven Smart Parking System is designed to enhance the efficiency of parking operations in metropolitan areas and high-traffic zones. By enabling drivers to swiftly identify available parking spaces, the system not only saves time but also alleviates traffic congestion. Furthermore, it promotes the optimal utilization of parking infrastructure, contributing to improved urban management. A dedicated IoT-integrated mobile application streamlines the process of locating vacant parking spots, thereby reducing search durations and enhancing user convenience. This initiative aims to mitigate urban parking challenges while fostering the development of sustainable transportation solutions. Within the scope of this study, a prototype system was developed, integrating hardware, embedded software, and a mobile application. Infrared (IR) sensors were employed for detecting parking space occupancy, while servo motors facilitated barrier control at entry and exit points. Arduino Mega served as the primary hardware controller, whereas NodeMCU, equipped with a Wi-Fi module, was utilized for data transmission via Firebase. The mobile application, developed using Flutter, provides functionalities such as user authentication, parking space monitoring, and reservation management. Additionally, the system allows for automated fee calculation based on usage duration, enabling payments through the application. To validate the system’s performance, various operational scenarios were tested, successfully demonstrating the functionality of both the prototype and the mobile application.