Decomposition Based Multicriteria Approach For Personalized Next-Basket Recommendation On Fashion Products

Hariyanto, Gilbert Kurniawan (2023) Decomposition Based Multicriteria Approach For Personalized Next-Basket Recommendation On Fashion Products. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 05111942000025-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 October 2025. Download (3MB) | Request a copy

Abstract

Fesyen dapat didefinisikan sebagai bentuk ekspresi diri dan otonomi pada periode tertentu dan tempat dan dalam konteks tertentu, pakaian, alas kaki, gaya hidup, aksesoris, makeup, gaya rambut, dan postur tubuh. Istilah tersebut menyiratkan tampilan yang didefinisikan oleh industri fashion sebagai tren. Karena maraknya produksi massal komoditas dan pakaian dengan harga lebih murah dan jangkauan global, kebutuhan media online untuk melakukan transaksi terkait industri fashion menjadi semakin penting. Meski demikian, masih banyak tantangan yang harus diselesaikan. Salah satunya mencakup variasi yang luas dari setiap barang yang tersedia untuk setiap merek yang kemudian menciptakan pilihan yang luar biasa bagi pelanggan untuk memilih mana yang mungkin merupakan pembelian yang cocok untuk setiap individu. Pada penelitian ini telah dilakukan pembuatan model Next-Basket Prediction dengan menggunakan metode Singular Value Decomposition (SVD) dan Sequential Pattern Analysis (SPA), dimana hasil akhir berupa prediksi dari 12 item yang akan dibeli oleh pelanggan. Hasil yang diperoleh kemudian dievaluasi menggunakan Mean Average Precision (MAP@12), dimana hasil yang lebih tinggi menunjukkan kinerja model yang lebih baik. Selama musim panas, SPA menunjukkan akurasi prediksi yang lebih tinggi daripada SVD, dengan nilai 0,00151242 dibandingkan dengan SVD 0,00136703. Di musim gugur, akurasi SVD meningkat menjadi 0,00167162, tetapi SPA terus mengungguli SVD dengan nilai 0,00245641, menunjukkan kekuatannya dalam menangkap pola berurutan. Peningkatan kinerja yang paling signifikan diamati pada musim dingin, di mana SVD mencapai nilai tertinggi secara keseluruhan pada 0,00215345, sementara SPA juga melampaui semua musim lainnya dengan nilai sebesar 0,00262222. Hasil ini menekankan keefektifan kedua metode dalam menangkap pola pembelian musiman dan memberikan prediksi yang akurat, terutama selama musim dingin.
==================================================================================================================================
Fashion can be defined as form of self-expression and autonomy at a particular period and place and in a specific context, of clothing, footwear, lifestyle, accessories, makeup, hairstyle, and body posture. The term implies a look defined by the fashion industry as that which is trending. Due to the rise of mass production of commodities and clothing at lower prices and global reach, the need of online media to do transaction related to fashion industry become more and more crucial. Despite of that, there are still numerous challenges needed to be solved. One of it includes the wide variety of each item available for each brand which then create an overwhelming choice for the customer to choose which may be a suitable purchase for each individual. In this research, the creation of a Next-Basket Prediction model has been carried out using the Singular Value Decomposition (SVD) and Sequential Pattern Analysis (SPA) methods, where the final output is a prediction of 12 items to be purchased by customers. The obtained results are then evaluated using Mean Average Precision (MAP@12), where higher results indicate better model performance. During the summer season, SPA showed higher prediction accuracy than SVD, with a value of 0.00151242 compared to SVD's 0.00136703. In the fall season, while SVD improved with value of 0.00167162, SPA continued to outperform SVD with a value of 0.00245641, indicating its strength in capturing sequential patterns. The most significant performance gains were observed in the winter season, where SVD achieved the highest value overall at 0.00215345, while SPA surpassed all other seasons with an outstanding value of 0.00262222. These results emphasize the effectiveness of both methods in capturing seasonal buying patterns and providing accurate predictions, especially during colder months.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Fashion, Next-Basket Recommendation, Recommendation System, Singular Value Decomposition, Sequential Pattern Analysis, Fesyen, Rekomendasi Keranjang Berikutnya, Sistem Rekomendasi
Subjects:	Q Science > QA Mathematics > QA76.9.I58 Recommender systems (Information filtering)
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55201-(S1) Undergraduate Thesis
Depositing User:	Gilbert Kurniawan Hariyanto
Date Deposited:	30 Jul 2023 07:42
Last Modified:	30 Jul 2023 07:42
URI:	http://repository.its.ac.id/id/eprint/99374

Actions (login required)

View Item