Coeus - Stardance

Ship #1 Pending review

@prathameshpvp on Coeus · about 3 hours ago

~ COEUS IS OUT!!
Coeus is a machine-learning application that explores how data science can help distinguish potential exoplanets from false-positive detections in NASA’s Kepler mission catalogue. Using ten astronomical and stellar features, including orbital period, transit duration, transit depth, planetary radius, and host-star properties. Coeus analyses patterns in Kepler Objects of Interest (KOIs) and predicts whether each object is classified as a planetary candidate or a false positive.

Built using Python and logistic regression, the project combines data preprocessing, predictive modelling, classification-threshold experiments, performance evaluation, and scientific visualisation. It investigates how measurable transit characteristics contribute to classification while examining

1 devlog
3h

Try project → See source code →

Open comments for this post

@prathameshpvp on Coeus · about 3 hours ago

3h 24m 4s logged

Coeus — Kepler Exoplanet Classification

Explored and cleaned NASA’s Kepler Objects of Interest (KOI) dataset.
Analysed 6,816 observations containing planetary candidates and false positives.
Selected 10 numerical features, including orbital period, transit duration, transit depth, planetary radius, and stellar properties.
Built a logistic regression model to classify planetary candidates and false positives.
Evaluated model performance using accuracy, precision, recall, F1-score, and a confusion matrix.
Investigated the effects of classification thresholds on candidate detection.
Visualised transit depth against orbital period to examine patterns and overlap between the two classes.
Identified limitations caused by class imbalance and overlapping feature distributions.
Worked on interpreting predictions cautiously, recognising that classification does not confirm the existence of an exoplanet.