This study looks at how individual socioeconomic factors relate to income levels among U.S. citizens, using a four-stage machine learning framework to piece things together. It started with prediction. We tested several regression models to estimate annual income based on features like education, employment status, debt, and household makeup. Each model brought something slightly different to the table, and together they helped sketch a clearer picture of the income landscape. Next came refinement. We dug into feature engineering, tuned the models, and brought in ensemble methods to pull out deeper patterns, especially the ones hiding in the interactions between things like education, housing, and digital access. In the third phase, we shifted focus to disparity. Using methods like ANOVA and t-tests, we looked at how income varies across groups, by race, gender, region, and marital status. The gaps were real and often held up even when we controlled for education or job type. The final step involved unsupervised clustering. This helped break the population into distinct socioeconomic profiles. Some clusters revealed vulnerable combinations, like high debt, spotty internet, and unstable work, that don’t always raise red flags on their own but matter when they show up together. What stood out through all of this is that income isn’t shaped by one factor at a time. It’s the result of how different parts of someone’s life overlap; region and education, debt and family structure, digital access and job opportunities. By combining prediction, diagnostics, and clustering, this approach gives both a close-up and wide-angle view of how income works. For researchers, it’s a way to move beyond surface-level forecasting. For policymakers, it offers a clearer path to spotting the groups most likely to fall through the cracks.