RAVEN: RAnking and Validation of ExoplaNets

We present RAVEN, a vetting and validation pipeline for TESS exoplanet candidates. RAVEN employs Gradient Boosted Decision Trees and Gaussian Process classifiers in a Bayesian framework to derive the posterior probability of a candidate being a planet against a set of False Positive (FP) scenarios. The classifiers are trained on synthetic training sets of planets and 8 astrophysical FP scenarios injected into TESS lightcurves, allowing for large-scale candidate vetting and performance verification against individual FP scenarios. A training set consisting of real TESS candidates caused primarily by stellar variability and systematic noise is also included. The machine learning probabilities are combined with scenario specific priors to compute the final posterior probabilities. These probabilities can be used for vetting candidates up to a radius of 16R⊕ and ranking them for follow-up. Candidates with probabilities greater than 99% against each FP scenario and whose implied radius is less than 8R⊕ are considered to be statistically validated by the pipeline. In this first version, RAVEN has been developed for candidates with a lightcurve from the TESS Science Processing Operations Centre, orbital period between 0.5 and 16 days and transit depth greater than 300ppm. RAVEN obtained area-under-curve (AUC) scores >97% on all FP scenarios and >99% on all but one. Testing on an independent external sample of 1361 pre-classified TOIs, the pipeline achieved an overall accuracy of 91%, with a precision of 97% and a recall score of 66%. RAVEN is publicly released as a cloud-hosted app, allowing easy access to the community.