Internet services are increasingly being deployed using cloud computing. However, the workload of an Internet service is not constant; therefore, the required cloud computing resources need to be allocated elastically to minimize the associated costs. Thus, this study proposes a proactive cloud resource scheduling framework. First, we propose a new workload prediction method—named the adaptive two-stage multi-neural network based on long short-term memory (LSTM)—which can adaptively route prediction tasks to the corresponding LSTM sub-model according to the workload change trend (i.e., uphill and downhill categories), in order to improve the predictive accuracy. To avoid the cost associated with manual labeling of the training data, the first-order gradient feature is used with the k-means algorithm to cluster and label the original training data set automatically into uphill and downhill training data sets. Then, based on stochastic queueing theory and the proposed prediction method, a maximum cloud service profit resource search algorithm based on the network workload prediction algorithm is proposed to identify a suitable number of virtual machines (VMs) in order to avoid delays in resource adjustment and increase the service profit. The experimental results demonstrate that the proposed proactive adaptive elastic resource scheduling framework can improve the workload prediction accuracy (MAPE: 0.0276, RMSE: 3.7085, $R^2$: 0.9522) and effectively allocate cloud resources.

With the rapid growth of distributed Internet of Things (IoT) services and edge-intelligence applications, conventional cloud computing is increasingly limited in latency-sensitive scenarios. Mobile Edge Computing (MEC) reduces latency by moving computation closer to end devices, while Unmanned Aerial Vehicles (UAVs) further extend MEC services to remote, emergency, or congested areas through flexible aerial deployment. However, UAV-enabled MEC still faces coupled challenges caused by heterogeneous tasks, limited resources, device energy constraints, time-varying channels, and UAV mobility. To address these challenges, this paper develops a Lyapunov-based joint optimization framework for UAV-enabled MEC systems. A dual-queue model is established to characterize local task uploading and UAV-MEC task execution, and a long-term stochastic energy minimization problem is formulated under queue-stability, resource-capacity, and energy constraints. By applying the drift-plus-penalty principle, the problem is transformed into online per-slot control decisions that jointly coordinate MEC scheduling, uplink power control, and UAV flight decisions. Structure-matched solutions are then developed, including a Lyapunov-drift-based MEC scheduling scheme, a queue-weighted closed-form water-filling power-control policy, and a gradient-based UAV flight controller with exponential smoothing and a fly-or-hover gate. A power–position alternating optimization algorithm is further introduced to handle the coupling between transmit power and UAV position. The simulation results demonstrate that the proposed framework maintains queue stability while reducing system energy consumption under heterogeneous and bursty workloads. It also achieves a balanced tradeoff between delay, energy consumption, and UAV flight activity, supporting energy-efficient and delay-aware UAV-MEC operation.