Integrasi konteks pengguna dan diversitas pada content-based filtering untuk rekomendasi resep makanan

Firdaus, Muhammad Rihap (2025) Integrasi konteks pengguna dan diversitas pada content-based filtering untuk rekomendasi resep makanan. Sarjana thesis, UIN Sunan Gunung Djati Bandung.

[img]
Preview
Text (COVER)
1_cover.pdf

Download (209kB) | Preview
[img]
Preview
Text (ABSTRAk)
2_abstrak.pdf

Download (250kB) | Preview
[img]
Preview
Text (SK BEBAS PLAGIARISM)
3_skbebasplagiarism.pdf

Download (165kB) | Preview
[img]
Preview
Text (DAFTAR ISI)
4_daftarisi.pdf

Download (244kB) | Preview
[img]
Preview
Text (BAB I)
5_bab1.pdf

Download (293kB) | Preview
[img] Text (BAB II)
6_bab2.pdf
Restricted to Registered users only

Download (460kB) | Request a copy
[img] Text (BAB III)
7_bab3.pdf
Restricted to Registered users only

Download (578kB) | Request a copy
[img] Text (BAB IV)
8_bab4.pdf
Restricted to Registered users only

Download (1MB) | Request a copy
[img] Text (BAB V)
9_bab5.pdf
Restricted to Registered users only

Download (256kB) | Request a copy
[img] Text (DAFTAR PUSTAKA)
10_daftarpustaka.pdf
Restricted to Registered users only

Download (227kB) | Request a copy
[img] Text (LAMPIRAN)
11_lampiran.pdf
Restricted to Repository staff only

Download (4MB) | Request a copy

Abstract

Fenomena information overload di era digital menyulitkan pengguna dalam menemukan informasi yang relevan secara cepat, termasuk dalam domain resep makanan. Melimpahnya platform dan konten resep daring sering kali membuat pengguna kewalahan karena dihadapkan pada terlalu banyak pilihan yang belum tentu sesuai dengan kebutuhan atau kondisi aktual mereka. Sistem rekomendasi menjadi solusi umum dengan menyaring informasi berdasarkan preferensi pengguna. Salah satu pendekatan yang banyak digunakan adalah Content-Based Filtering (CBF) karena sifatnya yang transparan dan independen terhadap data pengguna lain. Namun, CBF memiliki keterbatasan dalam menangkap konteks situasional dan cenderung menghasilkan rekomendasi yang homogen, yang berdampak pada turunnya kualitas pengalaman dan kepuasan pengguna. Penelitian ini bertujuan mengatasi keterbatasan tersebut dengan mengintegrasikan user-weighted context dan Maximal Marginal Relevance (MMR). Pendekatan user-weighted context menyesuaikan skor rekomendasi berdasarkan konteks eksplisit pengguna, sedangkan MMR digunakan untuk menjaga keseimbangan antara relevansi dan diversitas melalui metode reranking. Hasil evaluasi menunjukkan bahwa penerapan user-weighted context dan MMR secara signifikan meningkatkan performa rekomendasi pada masing-masing pendekatan dibandingkan baseline. Model yang diusulkan CBF + context + MMR memberikan kinerja paling seimbang dengan skor context satisfied@10 0.76 dari tertinggi 0.77 pada CBF + context, dibandingkan baseline 0.47 dan diversity@10 yakni 0.9 dari 0.68 pada baseline. Secara kualitatif, model ini menghasilkan rekomendasi yang lebih relevan dan bervariasi sesuai preferensi serta konteks pengguna, menjawab rumusan masalah dan menunjukkan bahwa integrasi konteks dan diversifikasi mampu meningkatkan kualitas sistem rekomendasi berbasis konten. The phenomenon of information overload in the digital era makes it difficult for users to quickly find relevant information, including in the domain of food recipes. The abundance of online recipe platforms and content often overwhelms users, who are faced with too many options that may not align with their actual needs or conditions. Recommender systems have become a common solution by filtering information based on user preferences. One widely used approach is Content-Based Filtering (CBF) due to its transparency and independence from other users' data. However, CBF has limitations in capturing situational context and tends to generate homogeneous recommendations, which negatively impacts user experience and satisfaction. This study aims to address these limitations by integrating user-weighted context and Maximal Marginal Relevance (MMR). The user-weighted context approach adjusts recommendation scores based on users' explicit context, while MMR maintains a balance between relevance and diversity through a reranking method. Evaluation results show that the application of user-weighted context and MMR significantly improves the recommendation performance of each approach compared to the baseline. The proposed model, CBF + context + MMR, delivers the most balanced performance, achieving a context satisfied@10 score of 0.76 (compared to a maximum of 0.77 in CBF + context and 0.47 in the baseline) and a diversity@10 score of 0.90 (compared to 0.68 in the baseline). Qualitatively, this model generates more relevant and varied recommendations aligned with user preferences and context, addressing the research problem and demonstrating that integrating contextual and diversification aspects can enhance the quality of content-based recommender systems.

Item Type: Thesis (Sarjana)
Uncontrolled Keywords: Sistem rekomendasi; Content-based Filtering; Context-aware; Diversitas; Maximal Marginal Relevance
Subjects: Data Processing, Computer Science > Computer Science Education
Special Computer Methods > Computer Pattern Recognition
Numerical Analysis > Algorithms
Divisions: Fakultas Sains dan Teknologi > Program Studi Teknik Informatika
Depositing User: Muhammad Rihap Firdaus
Date Deposited: 01 Sep 2025 03:56
Last Modified: 01 Sep 2025 03:56
URI: https://digilib.uinsgd.ac.id/id/eprint/116674

Actions (login required)

View Item View Item