Fashion is attached to people’s everyday life and their personality. Last decade fashion retailers intensely focused on personalized marketing to gain profit while letting customers reach the apparel they haven’t explored before. As a result of technological advancement, personalized recommendations came to play, which attracted many customers. Yet there are many areas to improve in fashion recommendation systems because users are looking for the ideal product for the money they spend. This paper discussed the drawbacks of current personalized recommendation systems and proposed solutions to implement a highly effective approach based on user preferences, purchase behavior, social factors, and fitting. The project successfully implements customer segmentation using K-means Clustering and extracting features from images using VGG16 pre-trained CNN with Keras library and Random Forest Method. Another fundamental idea of this project is to identify the body shapes of the users using CNN to understand the ideal fitting for fashion prediction. The future works of this project are to gather users’ actual data and use more advanced machine learning techniques to group customers by similarities.