This model predicts the probabilities of both weather-related Launch Commit Criteria (LCC) violations and launch attempts based on historical launch statistics and weather data. It aims to support launch scheduling decisions by providing insights into potential weather risks, aiding in the anticipation of LCC violations, and ultimately estimating the likelihood of a launch attempt.

Authors: Kyle Lyon, [email protected]

• Model date: March, 2024
• Model version: 1.0
• License: Apache 2.0

The workflow begins by loading and preprocessing historical launch statistics, launch forecasts, and hourly weather data. Feature engineering involves calculating average weather conditions during launch windows and joining this data with launch statistics and forecasts. The preprocessed data is then split into training and test sets. Various models, such as Logistic Regression or Random Forest, are trained and evaluated using metrics like accuracy, balanced accuracy, ROC AUC, and F1 score. The evaluation results are summarized in a Markdown table for comparison and analysis.

• Primary intended uses: This model is designed to assist the United States Space Force (USSF) in proactively identifying potential LCC violations and making informed launch scheduling decisions.
• Primary intended users: Personnel affiliated with AlgoByte LLC or the USSF, including launch weather officers, mission planners, and decision-makers.
• Out of scope use-cases: The models presented in this repository are strictly for research and demonstration purposes. They should not be solely relied upon to determine real-world space launch weather viability or LCC violation likelihood. Operational launch decisions should always be based on comprehensive weather assessments and established protocols.

• Data Dictionary:

• Source of training data: Eastern Range 5 SLS Launch Stats [Source]
• How training data was divided into training and validation data: 80% Training, 20% Validation
• Number of rows in training and validation data:
  • Training rows: 186, columns=16
  • Validation rows: 47, columns=16

• Source of test data: Eastern Range 5 SLS Launch Stats [Source]
• Number of rows in test data: 233
• State any differences in columns between training and test data: All the columns are as the same as the training & validation data.

• Columns used as inputs in the final model: visibility, dew_point, feels_like, temp_min, temp_max, pressure, humidity, wind_speed, wind_deg, clouds_all, rain_1h, rain_3h, weather_main, weather_description, weather_icon

• Column(s) used as target(s) in the final model: LAUNCHED

• Type of models:

  • Logistic Regression
  • Gradient Boosting
  • SVM

• Software used to implement the model: Python 3.10.13, Scikit-learn v1.4.1.post1.

• Coming later