Please use this identifier to cite or link to this item: http://repository.ipb.ac.id/handle/123456789/170293
Title: Pengembangan Sistem Rekomendasi Menggunakan Content-Based Filtering pada KMS Desa Digital
Other Titles: Development of a Recommendation System Using Content-Based Filtering on the KMS Desa Digital
Authors: Nurhadryani, Yani
Ahmad, Hafidlotul Fatimah
Ferdiansah, Muhammad Adi Satria
Issue Date: 2025
Publisher: IPB University
Abstract: Desa memiliki peran strategis dalam pembangunan nasional, namun banyak desa berkembang di Indonesia masih menghadapi tantangan dalam menemukan inovasi yang sesuai untuk diterapkan. Knowledge Management System (KMS) Desa Digital telah dikembangkan untuk mengelola dan menyebarkan berbagai inovasi desa, namun banyaknya pilihan dan tingginya indeks inovasi nasional menyulitkan pencarian yang relevan. Penelitian ini bertujuan mengembangkan sistem rekomendasi berbasis Content-Based Filtering (CBF) yang menyarankan inovasi berdasarkan kesamaan deskripsi inovasi yang telah diterapkan sebelumnya. Sistem diimplementasikan dalam KMS menggunakan FastAPI untuk backend, Firestore sebagai basis data, dan React untuk antarmuka pengguna. Evaluasi dilakukan dengan membandingkan hasil rekomendasi terhadap data ground-truth pada 10 inovasi uji, masing-masing dengan tiga item relevan, dan mengukur lima hasil rekomendasi teratas. Hasil menunjukkan rata-rata precision sebesar 0,36, recall 0,35, F1-score 0,35 dan Mean Average Precision (MAP) 0,48. Temuan ini menunjukkan bahwa sistem mampu mengidentifikasi sebagian inovasi relevan berdasarkan kemiripan deskripsi, meskipun masih terdapat keterbatasan akibat perbedaan persepsi semantik antara model dan pakar. Implikasi penelitian ini mengarah pada perlunya perluasan fitur representasi untuk meningkatkan akurasi relevansi inovasi di masa depan.
Villages play a strategic role in national development, yet many developing villages in Indonesia still face challenges in identifying suitable innovations for implementation. The Digital Village Knowledge Management System (KMS) has been developed to manage and disseminate various village innovations, but the abundance of available innovations and the increasing national innovation index make relevant discovery difficult. This study aims to develop a recommendation system based on Content-Based Filtering (CBF) that suggests innovations based on the similarity of descriptions to those already implemented. The system is integrated into the KMS using FastAPI for the backend, Firestore as the database, and React for the user interface. Evaluation was conducted by comparing recommendation results against ground-truth data for 10 test innovations, each with three relevant items, using the top five recommendations. The results show an average precision of 0.36, recall of 0.35, F1-score 0,35 and a Mean Average Precision (MAP) of 0.48. These findings indicate that the system can identify some relevant innovations based on textual similarity, although limitations remain due to semantic differences between the model’s approach and expert judgment. The implications of this study highlight the need for richer feature representations to improve the accuracy of innovation relevance in the future.
URI: http://repository.ipb.ac.id/handle/123456789/170293
Appears in Collections:UT - Computer Science

Files in This Item:
File Description SizeFormat 
cover_G6401211031_6ca8bddcf4e84b42adbc904f737a6eae.pdfCover2.33 MBAdobe PDFView/Open
fulltext_G6401211031_3451241d43be42dcba4c0403123d9614.pdf
  Restricted Access
Fulltext3.29 MBAdobe PDFView/Open
lampiran_G6401211031_2c41024faf8e49c982f6d5d5a13bc750.pdf
  Restricted Access
Lampiran2.14 MBAdobe PDFView/Open


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.