Identifying Exoplanets with Deep Learning: A CNN and RNN Classifier for Kepler DR25 and Candidate Vetting

The rapid expansion of exoplanet survey missions such as Kepler, TESS, and the upcoming PLATO mission has generated massive light-curve datasets that challenge traditional vetting pipelines. We introduce a hybrid deep-learning framework that integrates convolutional networks, bidirectional LSTMs, and an attention mechanism to identify planetary transit signals with improved accuracy and interpretability. Trained on Kepler DR25 data, the model achieves F1 = $0.910 \pm 0.008$ (AUC--ROC = $0.984 \pm 0.004$), significantly outperforming CNN-only baselines. Attention-based visualizations highlight ingress and egress phases consistent with astrophysical expectations. Applied to 1,360 DR25 candidate dispositions, our pipeline identified 190 high-confidence signals (P(model) $<$ 0.70) that passed initial validation. Following comprehensive false positive probability (FPP) analysis and contamination tests, 13 candidates achieved statistical validation, of which three were fully confirmed as robust exoplanets: KOI-901.01 (warm mini-Neptune), KOI-1066.01 (hot Jupiter), and KOI-212.01 (warm Neptune). With an inference time of $\sim 80$ ms per candidate, this framework enables scalable real-time triage for TESS and future PLATO operations, while the validated candidates provide a prioritized list for radial velocity, timing, and atmospheric follow-up. This interpretable, efficient pipeline addresses the bottleneck in exoplanet confirmation, advancing scalable discovery in the era of large-scale surveys.