Wireless networks with unmanned aerial vehicles (UAVs) as the airborne base station (ABS) become a promising technology to enhance terrestrial users’ coverage area. Moving at high speed over the wireless network covered by UAV requires an effective ABS selection scheme during the handover process to maintain the quality of service (QoS). On the other hand, the network load must be distributed fairly between the ABSs to prevent overloading at some base stations. Due to the limited power of ABSs, power saving is an essential task. Energy efficiency is a measure that balances data rate and power consumption. This study proposes a deep-predictive target ABS selection, resource block (RB), and power allocation algorithm that jointly maximizes energy efficiency and load balancing. The proposed method constructs the neighbor lists based on the reference signal received power (RSRP), signal to interference plus noise ratio (SINR), and the sojourn time estimated by the deep neural network model to reduce the unnecessary handovers. The simulation results show that the proposed method reduces the ping-pong rate and the outage probability and increases load balancing, spectral, and energy efficiencies compared to previous works.