FLP-XR

Official Python implementation of the FLP-XR model, proposed in the paper "FLP-XR: Future Location Prediction on Extreme Scale Maritime Data in Real-time”, submitted to IEEE Symposium on Maritime Informatics & Robotics, 2025.

Installation

In order to use FLP-XR, download all necessary modules in your directory of choice via pip, and install their corresponding dependencies, as the following commands suggest:

# Using pip/virtualenv
pip install −r requirements.txt

Data Preprocessing

In order to perform data preprocessing on the datasets used in the paper or your AIS dataset, run the following script inside the preprocessing folder.

sh run.sh

Inside the file main.rs, change the data path to accomodate your dataset, as well as any of the parameters (e.g. RATE, MAX_SPEED, STOP_SPEED_THR etc.) in order to fit your specific case.

Model usage and execution

Inside the model folder you will find the python code required to train and inference our model. In order to execute it, use the following command:

python model_train_inference.py 

while specifying all the required arguments, presented below:

-dataset_path       Path to dataset 
-dataset_name       Name of the dataset
-output_path        Path to save results
-sr                 Sampling rate of the dataset
-lr                 Learning rate of the model [DEFAULT: 0.01]
-max_depth          Maximum tree depth of the model [DEFAULT: 12]
-boosters           n_estimators, how many times to go through the boosting process, essentially how many trees to build. [DEFAULT: 750]