Gender data plays a crucial role in recommendation system for online shopping platforms, but issues like missing or inaccurate labels due to consumers’ reluctance to share information affect prediction accuracy. To address this, we analyze customer behavior from a vietnam FPT Group dataset, focusing on items viewed. Since the dataset is imbalanced, with female samples being one-third of male samples, here we segment the female samples into subsets and employ a two-layer classifier model for gender prediction. Our method achieves an average accuracy of 78% and completes predictions in few seconds. This lightweight approach efficiently handles gender prediction using machine learning algorithms, which offer valuable insights for decision-making in various applications.