This paper presents a novel content-based fashion recommendation system that leverages a deep ensemble classifier and transfer learning techniques. The system aims to enhance the accuracy and robustness of fashion image classification and recommendation by combining multiple pre-trained models, including MobileNet, DenseNet, Xception, and two variations of VGG. The probabilities obtained from these models are passed to a deep ensemble classifier, which uses cosine similarity to recommend similar products. The proposed method is evaluated using benchmark datasets such as the Fashion Product Images (Apparel) dataset and the Shoe dataset, demonstrating superior accuracy compared to existing models. The deep ensemble classifier achieves 96% accuracy, outperforming other models by up to 36%. The study highlights the potential of deep ensemble models in fashion image classification and recommendation systems, providing a practical solution for enhancing user experience in online fashion shopping.This paper presents a novel content-based fashion recommendation system that leverages a deep ensemble classifier and transfer learning techniques. The system aims to enhance the accuracy and robustness of fashion image classification and recommendation by combining multiple pre-trained models, including MobileNet, DenseNet, Xception, and two variations of VGG. The probabilities obtained from these models are passed to a deep ensemble classifier, which uses cosine similarity to recommend similar products. The proposed method is evaluated using benchmark datasets such as the Fashion Product Images (Apparel) dataset and the Shoe dataset, demonstrating superior accuracy compared to existing models. The deep ensemble classifier achieves 96% accuracy, outperforming other models by up to 36%. The study highlights the potential of deep ensemble models in fashion image classification and recommendation systems, providing a practical solution for enhancing user experience in online fashion shopping.