Analysis Pipeline for "Handover-Aware Conformal Throughput Intervals for Risk-Sensitive Rate Selection in Mobile 5G"

The work investigates whether conditioning conformal-prediction calibration on the mobility-regime state (radio-access technology × time since the last handover) improves throughput uncertainty estimates and the downstream rate-selection decision in mobile 5G.

Contents. Python scripts organized in nine numbered stages: data consolidation and strictly-causal feature construction; the base throughput predictor and residual diagnostics; the conformal-prediction variants (global, RAT-conditioned, and regime-conditioned, with a normalized-residual nonconformity score); the one-sided lower bound and trace-driven decision-layer emulation; multi-seed robustness runs; figure generation; sensitivity analyses; and the revision experiments (active-traffic/zero-inflation ablation, per-group calibration statistics, time-since-handover bucket sensitivity, a DtACI adaptive baseline, and an application-layer ABR/QoE emulation).