Shadin A Falcon 9 landings
IBM Data Science capstone
| Model | Validation | Test |
|---|---|---|
| Logistic regression | 84.6% | 83.3% |
| SVM | 84.8% | 83.3% |
| KNN | 84.8% | 83.3% |
| Decision tree | 88.9% | 66.7% |
The best validation score gave the worst test result.
- What ran
- A full pipeline predicting whether a SpaceX Falcon 9 first stage lands: launch data from the SpaceX API, cleaning, SQL and visual analysis, a Folium map of launch sites, a Plotly Dash dashboard, and four tuned classifiers (logistic regression, SVM, decision tree, KNN).
- What worked
- Logistic regression, SVM and KNN each predicted 83.3% of the held-out test launches correctly. The decision tree looked best in cross-validation at 88.9%, then dropped to 66.7% on the test set, so the validation score alone would have picked the wrong model.