A trajectory prediction model based on deep feature representation

The core module of this model architecture focuses on processing subtrajectories to capture complex movement patterns:

: Analyzes the timing and duration of movements within the sliding window.

: A secondary feature where a self-attention model assigns weights to these extracted features to prioritize the most relevant data points for predicting future longitude and latitude.

A primary "deep feature" of this system is its ability to extract from trajectory data through a dedicated representation layer. Key Deep Feature: Multi-Dimensional Extraction

: Captures the geometric and geographic positioning of a path.