MISA-XGModel 0.2.2

A Python library for predicting electron density using XGBoost

MISA_XGModel

A Python library for predicting electron density using XGBoost.

---

Overview

This library provides tools to predict electron density (Ne) using a machine-learning model based on XGBoost. It supports querying predictions for specific latitude, longitude, day of year (DOY), altitude, and solar local time (SLT), while leveraging pre-computed geophysical indices. The library also supports efficient vectorized predictions for multiple inputs.

---

Features

• Predict electron density (Ne) for specific geospatial and temporal conditions.
• Efficient batch predictions using vectorized input arrays for higher performance.
• Clamp or wrap input values to the bounds of the training dataset for robustness.
• Supports querying geophysical indices from provided datasets.
• Modular design for integrating geophysical models in Python applications.

---

Installation

You can install this package via pip:

pip install MISA_XGModel

The model downloads missing dependencies on first import (e.g., the XGBoost model weights, scaler, and geophysical dataset):

Downloading model from https://www.dropbox.com/...
Model downloaded and saved to data/xgboost_optimized_model.json.
Downloading scaler from https://www.dropbox.com/...
Scaler downloaded and saved to data/scaler_large.json.
Downloading dataset from https://www.dropbox.com/...
Dataset downloaded and saved to data/master_geo_ds.nc.

If the files already exist locally, the library will skip the download step and use the existing files.

---

Usage

Quickstart

Here’s how you can use the library to predict electron density:

Single Prediction

from MISA_XGModel import predict_ne, query_model

# Predict Ne with dataset lookup for geophysical indices
predicted_ne = predict_ne(
    lat=42.0, lon=-71.0, doy=99, alt=150.0, slt=12.0, year=2024
)
print(f"Predicted Ne: {predicted_ne:.2e}")

# Predict Ne with precomputed geophysical indices
predicted_ne = query_model(
    lat=42.0, lon=-71.0, doy=99, alt=150.0, slt=12.0,
    hp30=2, ap30=7, f107=209, kp=2.3, fism2=0.0007678
)
print(f"Predicted Ne: {predicted_ne:.2e}")

Batch Predictions

To predict multiple inputs efficiently, you can pass arrays to predict_ne or query_model:

import numpy as np

# Vectorized inputs
lats = np.array([42.0, 41.5, 40.0])
lons = np.array([-71.0, -72.0, -73.0])
doys = np.array([99, 100, 101])
alts = np.array([150.0, 200.0, 250.0])
slts = np.array([12.0, 14.0, 16.0])
year = 2024

# Batch predict Ne with dataset lookup for geophysical indices
predicted_ne = predict_ne(
    lat=lats, lon=lons, doy=doys, alt=alts, slt=slts, year=year
)
print(f"Predicted Ne: {predicted_ne}")

# Batch predict Ne with precomputed geophysical indices
predicted_ne = query_model(
    lat=lats, lon=lons, doy=doys, alt=alts, slt=slts,
    hp30=np.array([2, 3, 4]),
    ap30=np.array([7, 8, 9]),
    f107=np.array([209, 210, 211]),
    kp=np.array([2.3, 2.5, 2.7]),
    fism2=np.array([0.0007678, 0.00078, 0.00079])
)
print(f"Predicted Ne: {predicted_ne}")

---

Inputs and Parameters

Clamping and Wrapping Input Values

Input parameters (lat, lon, alt, slt) are clamped to the boundaries of the training data, while doy is wrapped to stay within [0, 364]. These boundaries are defined as follows:

Parameter	Min Value	Max Value	Notes
lat	37.5	49.9	Clamped to range
lon	-85.7	-76.1	Clamped to range
alt	94.6 km	500 km	Clamped to range
slt	0.0 hrs	24 hrs	Clamped to range
doy	0	364	Wrapped (e.g., 365 → 0, -1 → 364)

---

Requirements

• Python 3.7+
• Dependencies:
  • xgboost
  • scikit-learn
  • numpy
  • pandas
  • xarray
  • joblib
  • netcdf4

---

License

This project is licensed under the MIT License - see the LICENSE file for details.

---

Author

• Mateo Cardona Serrano (them)
• GitHub Profile
• Email

---

Acknowledgments

• Thank you to Sevag Derghazarian for his continuous support and consultation on this project.

---