MESSENGER Magnetospheric Classification Using LSTM Neural Network

Overview

This repository hosts Python scripts for supervised learning using an LSTM neural network to classify magnetospheric data from the MESSENGER spacecraft. The methodology and findings are detailed in (Smith, Jackman et al. 2023). Data used included MESSENGER magnetomster data, and manually labeled magnetospheric crossings used for training are available in (Sun et al. 2020).

Model Training

To train the LSTM neural network model with the MESSENGER mission data, follow these steps:

1. Data Preprocessing

   • Run the preprocessing script located at data_prep/messenger_preprocessing_pipeline.py.
   • This script processes the MESSENGER CSVs and Sun labels for model training.
   • Be sure to include the relevant data files in the data directory.

2. Model Training

   • Use the RNN training script located at ml_models/messenger_rnn.py.
   • This script trains the LSTM neural network model with the preprocessed data.

3. Result Visualisation

   • Visualise the training results using the script ml_models/visuals.py.
   • This script aids in visualising the outcomes as included in the associated research paper.

Using the Model for New Labels

Getting Started

Prerequisites

• Python 3.x
• Required Python packages (see requirements.txt)

Installation

Clone the repository:

git clone https://github.com/ksmith9/messenger_ml_mag_classification.git