In Wireless Sensor Networks, diverse nodes are associated with each other for monitoring definite circumstances. So, sensors are considerably utilized in distinct real-time utilizations namely remote operated unmanned vehicle, atmospheric surveillance, disaster management, and so on. Transmitting data from a remote operated unmanned vehicle to server via Long Term Evolution (LTE) with the harmony of Bluetooth Low Energy (BLE) relaying remains the core of significant data transmission in wireless networks. The utilization of Unmanned Aerial Vehicles (UAVs) for wireless networks is swiftly heightening as the driving force of new applications due to their distinctive resources for improving coverage and energy efficiency of wireless network UAVs act as base stations. In other condition, data-driven Deep Learning-assisted (DL) strategies using multilayer perceptron are acquiring an increasing interest for not utilizing huge frequency of generated data, however ensuring network procedure in an optimal manner and hence providing QoS requirements of wireless networks. But, UAVs is resource-constrained devices specifically in power resources and data transmission. With traditional DL scheme being cloud-centric necessitate UAVs' data are stored in centralized server, therefore generating huge communication overhead and thus result in network bandwidth and energy inefficiency of UAV devices. To address these issues in this work, a Fully Recursive Long Short Term Memory (FR-LSTM) for improving data transmission rates and quality of service in wireless networks is proposed. Initially, Deep Learning-based model was designed in Long Term Evolution (LTE) Dominant Influencing Criterions (DIC) estimation. The applications of power resources and bandwidth allocation (PRBA) in self-organizing LTE small cell network, therefore minimizing RMSE and average end-to-end delay involved in transmission. Next, a Fully Recursive Perceptron Network (FRPC) and LSTM model was utilized and applied for DIC to resolve the UAV position which reduces overall system performance and user throughput. Hence, for classification regression tasks, when is there no LTE signal, data can be transmitted to another device through BLE (Bluetooth Low Energy), therefore ensuring throughput and ensuring minimum latency. The effectiveness of FR-LSTM is yet to be validated using four kinds of evaluation metrics with diverse number of nodes, namely, RMSE, throughput, average end-to-end delay, and latency.