UNDERSTANDING THE EXOPLANET DETECTION SYSTEM
This system leverages a Random Forest Classifier trained on the NASA Kepler Exoplanet dataset. It analyzes key telemetry data—such as orbital period, transit depth, and signal-to-noise ratio—to distinguish between Confirmed Exoplanets and False Positives. By processing 7 critical features and engineering derived metrics like insolation flux, the system predicts the likelihood of a planet being a exoplanet with high accuracy.
Raw telemetry data is ingested from the NASA Kepler Exoplanet dataset, containing over 9,000 recorded objects of interest (KOIs).
The FastAPI backend calculates complex derived metrics—such as log_insolation and radius ratios—from the raw user inputs to match the model's training parameters.
Our pre-trained Random Forest Classifier (200 estimators) analyzes the engineered features to output a binary classification with a confidence score.
Understanding the data points used by the System to make its decision.
The time it takes for the planet to complete one full orbit around its star. (A "Year" on that planet).
VisualizeHow much starlight is blocked during the eclipse, measured in Parts Per Million. Deeper dips usually mean bigger planets.
VisualizeHow many hours the planet spends crossing the face of the star. Helps determine the orbit speed and angle.
VisualizeSignal-to-Noise Ratio. A higher number means the detection is very clear and distinct from background static.
VisualizeThe estimated size of the planet compared to Earth. (e.g., 1.0 = Earth Size, 11.2 = Jupiter Size).
Compare SizeThe size of the star compared to our Sun. We need this to calculate the planet's relative size accurately.
Compare StarHow much energy/heat the planet receives from its star. High flux = hot roasting planet; Low flux = icy world.
Visualize Heat