Abstract—This paper presents a novel approach to modeling the attenuation of millimeter wave (mmWave) signals, using deep learning techniques using IoT sensor data from the University of Mosul. This paper aims to significantly improve prediction accuracy under various environmental conditions, such as water vapor, oxygen, and rain. The research shows that combining Convolutional Neural Networks (CNN) with Recurrent Neural Networks (RNN) leads to a significant improvement in predicting signal attenuation that outperforms traditional models. The paper also discusses the integration of IoT with 5G using deep learning to analyze pollutant data to provide essential tools for the development of smart cities. These deep learning models excel at capturing complex nonlinear environmental interactions, covering more reliable mmWave signal attenuation predictions. The results show that dust could have a good side for spectral efficiency because of the ability to increase the frequency reuse factor in cellular systems. This insight paves the way for future research to explore the effect of dust on spectral efficiency, expanding the focus beyond the mere attenuation and visibility.