You are working on enhancing an ecommerce platform by developing a deep learning model to predict the likelihood of a customer making a purchase. After evaluating the performance of your model using both the original training data and a new set of test data, you observe that your model suffers from overfitting. This means the model performs well on the training data but poorly on the test data, indicating that it does not generalize well to unseen data. Your objective is to improve the model's accuracy for new, unseen data. What steps should you take to address this issue?