This data science project for CST383: Introduction to Data Science investigates the relationship between a satellite’s physical characteristics and its intended mission. By developing a predictive model to categorize satellites into six primary classes, my team and I demonstrated that specific mission types are strongly tied to distinct orbital patterns. Key achievements include:
- High Accuracy Modeling: Developing a predictive model using multinomial Logistic Regression that achieved a 94.4% accuracy rate.
- Feature Engineering: Analyzing key orbital parameters such as inclination, apogee, and eccentricity alongside physical features like launch mass.
- Data Integration: Utilizing and merging comprehensive datasets from the UCS Satellite Database and Celestrak SatCat.
- Actionable Insights: Demonstrating that publicly available orbital data serves as a powerful proxy for identifying the purpose of unknown satellites.
The research illustrates a scalable method for rapid cataloging and provides significant engineering insight into the deployment of orbital assets. Complex behaviors like orbital missions leave patterns in data that can be decoded with the right statistical tools.