Anzary, Alif Adrian (2024) Pemilihan Influencer: Integrasi Citra dan Teks Menggunakan Attention-Based Neural Network Untuk Pemasaran Brand. Other thesis, Institut Teknologi Sepuluh Nopember.
Text
5025201274-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 October 2026. Download (8MB) | Request a copy |
Abstract
Tugas Akhir ini menggunakan metode Optical Character Recognition (OCR), Attention Based Neural Network dan Learning to Rank (LTR) untuk memilih influencer Instagram yang sesuai untuk kampanye pemasaran brand. Dengan fokus pada pengenalan karakter optik dari konten citra dan teks, penelitian ini mengembangkan model untuk otomatisasi seleksi influencer dengan mempertimbangkan relevansi konten, engagement rate, dan citra brand. Analisis dilakukan pada data Instagram untuk menguji efektivitas model dalam mencocokkan brand dengan influencer yang paling sesuai.
Pada Tugas Akhir ini, dilakukan prediksi ranking influencer untuk suatu brand dengan dataset profile influencer dan brand dari sosial media Instagram. Tahapan yang dilakukan pada Tugas Akhir ini adalah Persiapan data, praproses data, penghitungan skor relevansi, pembuatan model dan pengujian model menggunakan metrik MRR, MAP, dan NDCG. Hasil akhir pada penelitian ini adalah rekomendasi lima dan sepuluh influencer teratas untuk suatu brand.
Berdasarkan pengujian evaluasi dataset 38.113 influencer Instagram profile dan 25.282 brand Instagram profile, hasil rekomendasi influencer untuk suatu brand model kinerja paling baik ketika dilatih adalah Pointwise menggunakan fungsi kerugian Squared Loss Function. Model ini mencapai waktu tercepat 0,0020 detik, menunjukkan efisiensi tinggi dalam Persiapan dan rekomendasi. Hasil evaluasi menunjukkan bahwa model ini menghasilkan metrik MRR@5 dan MRR@10 sebesar 1,0000, serta metrik MAP@5 dan MAP@10 sebesar 1,0000, menandakan kemampuannya dalam memberikan rekomendasi yang tepat dan relevan.
===========================================================
This paper uses Optical Character Recognition (OCR), Attention Based Neural Network and Learning to Rank (LTR) methods to select suitable Instagram influencers for brand marketing campaigns. With a focus on optical character recognition of image and text content, this research develops a model for influencer selection automation considering content relevance, engagement rate, and brand image. Analysis is conducted on Instagram data to test the effectiveness of the model in matching brands with the most suitable influencers.
In this Final Project, the prediction of influencer ranking for a brand is carried out with a dataset of influencer and brand profiles from Instagram social media. The stages carried out in this Final Project are data processing, data extraction and preprocessing, calculation of relevance scores, model building and model testing using MRR, MAP, and NDCG metrics. The final result of this research is the recommendation of the top five and top ten influencers for a brand.
Based on the evaluation testing of a dataset of 38.113 influencer Instagram profiles and 25.282 brand Instagram profiles, the best performing influencer recommendation for a brand model when trained is Pointwise using the Squared Loss Function. This model achieved the fastest time of 0,0020 second, showing high efficiency in processing and recommendation. The evaluation results show that this model produces MRR@5 and MRR@10 metrics of 1,0000, as well as MAP@5 and MAP@10 metrics of 1,0000, signifying its ability to provide appropriate and relevant recommendations.
Item Type: | Thesis (Other) |
---|---|
Uncontrolled Keywords: | Brand, Influencer, Instagram, Learning to Rank, Optical Character Recognition. |
Subjects: | T Technology > T Technology (General) > T57.5 Data Processing |
Divisions: | Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55201-(S1) Undergraduate Thesis |
Depositing User: | Alif Adrian Anzary |
Date Deposited: | 01 Aug 2024 22:30 |
Last Modified: | 01 Aug 2024 22:30 |
URI: | http://repository.its.ac.id/id/eprint/111061 |
Actions (login required)
View Item |