Background: Predicting customers’ income can be a powerful tool in helping retailers gain valuable insights and personalize their income-based marketing strategies, especially in a highly competitive market like the retail industry. Much research has used machine learning to predict income levels in different contexts. However, our research stands out by applying algorithms specifically to the retailer using k-fold cross-validation to estimate Machine Learning models with two approaches: Feature Selection and Direct Apply. Aim: Our research aims to explore consumer trends by analyzing available data and applying techniques to predict consumers’ income levels. Method: Multiple Linear Regression, Bayesian Ridge Regression, and Random Forest Regression were selected for prediction. Moreover, k-fold cross-validation is used for Machine Learning effort estimates in Direct Apply and Feature Selection while being evaluated through five evaluation criteria (MAE, MAPE, PRED (0.25), RMSE, and Paired T-test). SHAP analysis was also employed to interpret the results of various machine learning models. Results: The key consumer demographic and shopping preferences were identified. Regarding prediction, Feature Selection performed better than direct application, and Random Forest Regression achieved the highest accuracy in general. The Paired T-test analysis indicates a statistically insignificant difference between predicted and actual values in six models. SHAP analysis revealed that Random Forest Regression outperformed the other regression models by effectively capturing complex feature interactions and non-linearities. Conclusion: Those valuable insights provide an overview of customers’ portraits and their behavior trends based on income. Machine learning algorithms are encouraged to be used to predict individual income levels in retail selling and allow businesses to tailor strategies to promote customer loyalty.