The model was trained on ~5000 rows of data from the 3rd campaign of the K2 mission. It uses a classification algorithm to detect probability of a transit. Currently, the model is able to produce:
- 100% TP
- 89.6% TN
- 10.4% FP
- 0% FN
Currently, the model used to classify exo-planet transits in this app is the XGBoost Classifier. This model was chosen due to the need for recursive building of new trees with prior knowledge. Because the data contains so few positive classes, the few positives that do exist were oversampled, and the model is used specifically to identify non-transit systems, with a positive class indicated by just 1% probability of a transit.
The app is hosted here using Heroku. Please allow up to 1 minute for the app to "wake up".