About the dataset: A synthetic heart-disease dataset built for 2026: HbA1c, full lipid panel, wearable step counts, sleep and stress - the risk factors the classic 300-row heart datasets never recorded. Ideas to get you started: - Build a heart-disease classifier and check which factors carry it. - Do wearable owners actually show a lower risk profile? Test it properly - and watch for confounding by age. - Which matters more: the lipid panel or the lifestyle columns? Compare model families. - Is the sleep-risk relationship really U-shaped? Bin sleep_hours and plot it. - Build a simple interpretable risk score and compare it to a gradient-boosted model. - Explore HDL vs LDL - does the ratio beat either one alone?