Asteroid Diameter Prediction

This project leverages machine learning to predict asteroid diameters using astronomical data. The dataset, sourced from Kaggle, includes features like semi-major axis, eccentricity, reflectivity (albedo), and orbital characteristics. The analysis benchmarks results against prior work utilizing machine learning models like Random Forest, LightGBM, and XGBoost..

Key steps in the project include:

• Exploratory Data Analysis (EDA): Insights reveal that features like Minimum Orbit Intersection Distance (MOID) and semi-major axis exhibit weak to moderate correlation with asteroid diameter. Clustering techniques like K-Means are suggested for capturing nuanced relationships.
• Model Performance: Random Forest Regressor performed the best with an R² score of 0.9619, followed by LightGBM and XGBoost. Feature importance analysis highlighted key predictors of asteroid diameter.
• Evaluation Metrics: Models were assessed using R², Adjusted R², RMSE, and MAE. Data preprocessing and feature engineering significantly improved performance compared to baseline results from the referenced research.

The project demonstrates the potential of machine learning in space science while highlighting data challenges and modeling nuances.