- NebulaNet is a cutting-edge data science hackathon organized by the physics department of IIT BHU for Jigyasa. Participants were provided with a Dataset containing astrophysical data. They had to train a Machine Learning Model which optimizes the given dataset, focusing on how ml techniques can be applied to physics research.
| Column ID | Column Name | Data type | Values type | Description |
|---|---|---|---|---|
| 0 | Temperature (K) | int64 | Continous | Temperature of stars |
| 1 | Luminosity(L/Lo) | float64 | Continous | Luminosity of stars |
| 2 | Radius(R/Ro) | float64 | Continous | Radius of stars |
| 3 | Absolute magnitude(Mv) | float64 | continous | Magnitude of stars |
| 4 | Star type | object | Discrete | Types of stars |
| 5 | Star color | object | Discrete | Colours of stars |
| 6 | Spectral Class | object | Discrete | Spectral class of stars |
- It contains categorical and numerical variables.
- It is skewed and contains outliers.
- It is not normally distributed.
- Relationships between various variables has been analysed and significant points have been highlighted.
- While working on the project, I discovered various physics terms & formulaes which have impact on star type.
- Random Forest Classifier , Decision Tree Classifier & Xgboost get 1.00 as accuracy score. We select Random Forest Classifier.The reason for the same is that using bagging algorithms is much better than using boosting algorithms for a small dataset.