Ahmadi, Fadias Nur (2026) Integrasi Model Arsitektur Nlp Transformer dan K-nearest Neighbour Dalam Rekomendasi Busana: Sistem Pendukung Keputusan Outfit Matching. Undergraduate thesis, Politeknik Negeri Jember.
![[img]](https://sipora.polije.ac.id/style/images/fileicons/text.png) |
Text (Abstract)
Abstract.pdf - Submitted Version Available under License Creative Commons Attribution Share Alike. Download (7kB) |
|
Text (Bab 1 Pendahuluan)
Bab 1 Pendahuluan.pdf - Submitted Version Available under License Creative Commons Attribution Share Alike. Download (143kB) |
|
Text (Daftar Pustaka)
Daftar Pustaka.pdf - Submitted Version Available under License Creative Commons Attribution Share Alike. Download (139kB) |
|
Text (Laporan Lengkap)
Laporan Lengkap.pdf - Submitted Version Restricted to Registered users only Download (3MB) | Request a copy |
Abstract
The rapid development of the digital fashion industry has increased the demand for clothing recommendation systems that are more personalized and adaptive. However, existing recommendation systems generally focus on visual similarity or user prefer- ences without considering users' physical characteristics, such as body proportions and skin tone. This study proposes an Artificial Intelligence (AI)-based outfit recom- mendation system by integrating the Multimodal Transformer model (OpenAI CLIP), the K-Nearest Neighbors (KNN) algorithm, and an Expert System. The CLIP model is utilized to extract visual and semantic features, while KNN is employed to retrieve out- fit candidates with the highest similarity. Furthermore, the Expert System incorporates Body Mass Index (BMI) and skin tone categories based on the Individual Typology Angle (ITA) concept to generate more personalized outfit recommendations. The da- taset used in this study consists of 12,500 filtered outfit samples from the Polyvore Dataset. The evaluation results show that the CLIP model achieved a classification accuracy of 60% across four fashion style categories: Casual, Formal, Smart Casual, and Sporty. The proposed system is capable of generating the top five outfit recommen- dations along with styling suggestions tailored to the user's physical characteristics.
Actions (login required)
 |
View Item |
