Modified Long Short Term Memory-Puzzle Optimization Algorithm for Task Offloading Prediction in Mobile Edge Computing

Task offloading may increase transmission latency as well as transmission energy use in IoT edge computing. The paper describes the task offloading problem as a joint decision-making problem for cost reduction that also encompasses processing energy consumption. Processing latency, and the task throw rate associated with latency-sensitive tasks in order to lower the cost of resources needed for task offloading and enhance the utilization of server resources. The MLSTM networks-based POA algorithm is suggested to resolve the aforementioned task offloading decision dilemma. This approach uses the MLSTM method to anticipate the load on the edge server in real-time during the model’s training stage, which significantly enhances speed and convergence accuracy throughout the offloading procedure. In the testing stage, the next task’s characteristics are predicted using an MLSTM framework, and next the task’s computing resources are assigned, thereby decreasing the task’s reaction time and improving the system’s offloading effectiveness. According to the experimental assessment, this approach may successfully lower offloading costs, average latency, and task throw rates.