Deep Mobile Path Prediction with Shift-and-Join and Carry-Ahead

Importance of user mobility has rapidly increased in 5G due to reduced cell sizes, management Multi-access Edge Computing (MEC), and ultra-low latency services. Reactive nature existing systems is a bottleneck, it can be solved by building proactive that exploit temporal characteristics time-series data predict long-term movement (i.e.path). However, mobile path prediction with useable accuracy challenging task, particularly for lengthy target trajectories. This paper adopts general approaches propose two models predicting high accuracy. Step Forward Iteration (SFI) model based on recursive approach, whereas Encoder-Decoder (ED) follows multi-output both the use Long-Short Term Memory (LSTM) as learning unit. Training testing these done datasets from wireless network Pangyo ICT Research Center, Korea one Korean operators. The experiment results show viability proposed leveraging management, they outperform state-of-the-art GRU attention (GRU-ATTN) Transformer Network (TN) models. highest accuracies achieved 3, 5, 7 steps sequences (i.e.predicted path) campus dataset are 96%, 90%, 87%, respectively.