Fashion Recommendation System Using Machine Learning And CNN: Simulation-Based Approach

Abstract

Fashion recommendation systems are critical in enhancing user experience on e-commerce platforms. This study presents a machine learning-based framework employing collaborative filtering, content-based filtering, and hybrid models, enhanced by Convolutional Neural Networks (CNNs) for image-based recommendation. We simulated algorithms including K-Nearest Neighbors (KNN), Support Vector Machine (SVM), Random Forest, and CNN using the Fashion Product Images dataset. Evaluation through accuracy, precision, recall, and F1-score demonstrates that CNN-based hybrid recommendations significantly improve performance. This paper proposes a hybrid machine learning (ML) system for personalized fashion recommendations, integrating visual content-based filtering (using deep learning) and collaborative filtering to address cold-start and scalability challenges. Our framework processes multimodal data (images, text, user behavior) using a ResNet-50 CNN for image feature extraction, BERT for text embeddings, and matrix factorization for collaborative signals