The paper explores the potential for machine learning techniques to identify housing units at high risk of gentrification in Washington D.C. Metropolitan Statistical Area, using longitudinal aspect of American Housing Survey data. It detects housing units with new residents with much higher income and finetunes the gentrification cases adding more filters with questions about the reasons to move. After implementing several ML models, Random Forest Classifier model is adopted for the best prediction model which yields 83 percent accuracy, 81 percent prediction, and 87 percent recall score. Further analysis on the profile of gentrifiers and feature importance suggests that highly educated young adults looking for apartments in walkable urban neighborhoods have driven the gentrification in Washington D.C. MSA.