Fusion: Practice and Applications FPA 2692-4048 2770-0070 10.54216/FPA https://www.americaspg.com/journals/show/2099 2018 2018 Scientific Research Center, Al-ayen University, Thi-Qar, Iraq Sajad Ali Zearah Department of Medical Devices Engineering Technologies, National University of Science and Technology, Dhi Qar, Nasiriyah, Iraq Ahmed R. Hassan Medical instruments engineering techniques, Al-farahidi University, Baghdad, Iraq Aqeel Ali Computer Technologies Engineering, Al-Turath University College, Baghdad, Iraq Saad Qasim Abbas Technical Computer Engineering Department, Al-Kunooze University College, Basrah, Iraq Tamarah Alaa Diame Department of Computer Engineering, University of Massachusetts Dartmouth, MA 02747Inst, USA Ahmed Mollah Khan Department of Computer Science and Engineering, University of Deusto, 48007 Bilbao, Spain Mariok Jojoal Device-to-Device (D2D) Communication promises outstanding data speeds, overall system capacity, and spectrum and energy efficiency without base stations and conventional network infrastructures, and these improvements in network performance sparked a lot of D2D research that exposed substantial challenges before being used to their fullest extent in 5G networks. This study suggests using Deep Learning-based Improved D2D communication (DLID2DC) in 5G networks to address these issues. Reprocessing resources between Cellular User Equipment (CUE) and D2D User Equipment (DUE) can increase system capacity without endangering the CUEs. The D2D resource allocation method allows for a flexible distribution of available resources across CUEs. In addition, several CUEs can consume the same pool of resources simultaneously. Researchers utilize various deep learning techniques to handle the difficulty of constructing D2D links and addressing their interference, mainly when using millimeter-wave (mmWave), to improve the performance of D2D networks. This research aims to increase system capacity by optimizing resource allocation using the suggested DLID2DC paradigm. The model uses Deep Learning methods to overcome interference issues and make D2D link building more efficient, especially in mmWave communication. The model uses Convolutional Neural Networks (CNNs) to learn and adapt to complicated D2D communication patterns, improving performance and dependability. The experimental findings show that, compared to other conventional approaches, the proposed DLID2DC model improves connection with lower end-to-end delay, energy efficiency, throughput, and efficient convergence time. 2023 2023 203 220 10.54216/FPA.130116 https://www.americaspg.com/articleinfo/3/show/2099