An-Najah Staff

Publication Type

Original research

Authors

Fulltext

Download

During recent years, the accelerated evolution of online fashion trends and seasonal

dynamics has increased the demand for intelligent recommendation systems capable

of balancing personalization and novelty. This study presents a hybrid

artificial-intelligence-based fashion recommender system that generates personalized

and trend-aware clothing suggestions based on the contents of users’ virtual closets.

The proposed model integrates item-based collaborative filtering (IBCF) and

content-based filting (CBF), augmented by a Sugeno fuzzy-logic module that

computes a continuous trend percentage (T) using normalized sales, likes, and views.

Experiments were conducted on a dataset of 4,224 items using simulated virtual

closet ranging from 40 to 80 items. Performance was evaluated using precision at

rank 10 (P@10), recall at rank 10 (R@10), and Normalized Discounted Cumulative

Gain at rank 10 (NDCG@10), in addition to runtime scalability vs. catalog size. The

proposed fuzzy-hybrid model consistently outperformed pure content-based, pure

collaborative, and non-fuzzy hybrid baselines in recommendation accuracy while

maintaining linear scalability. The results indicate that fuzzy-logic-based trend

modeling effectively enhances the balance between stylistic consistency and exposure

to emerging fashion trends within the evaluated experimental setting. However, the

observed improvements should be interpreted cautiously because the experiments

rely on simulated user closets and a private dataset. The findings suggest that the

proposed approach can inform the design of adaptive fashion recommendation

systems. However, results are obtained under a controlled simulation setting with

manually designed fuzzy membership functions, and real-user validation remains an

avenue for future work.

Journal