Analisis Fitur Lagu Yang Mempengaruhi Jumlah Pemirsa di Youtube Menggunakan Metoda Machine Learning dan Librosa Features Extraction

Wijayanto, Unggul Widodo (2019) Analisis Fitur Lagu Yang Mempengaruhi Jumlah Pemirsa di Youtube Menggunakan Metoda Machine Learning dan Librosa Features Extraction. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 05111540000154-Undergraduate_Theses.pdf Restricted to Repository staff only Download (1MB) | Request a copy

Abstract

Perkembangan teknologi sangatlah pesat. Salah satunya adalah dalam bertukar video. Youtube merupakan website pertukaran video yang populer. Salah satu kategori yang sering diunggah adalah musik. Produser musik memprediksi tingkat kesuksesan yang dapat diukur dengan menggunakan jumlah views. Tugas akhir ini memprediksi popularitas lagu dengan diukur jumlah views-nya dengan membandingkan antara fitur fundamental dan teknikal. Fitur Fundamental merupakan karakteristik internal sedangkan teknikal merupakan karakteristik eksternal musik tersebut. Penulis menggunakan video musik pop sebagai dasar analisis. Pertama pembangunan dataset. Kedua, ekstraksi fitur fundamental dan teknikal. Ketiga, praklasifikasi fitur-fitur dengan menggunakan Stratified K-Fold, Normalisasi, seleksi fitur CHI2 atau Random Forest, dan oversampling. Keempat, melakukan klasifikasi dengan menggunakan machine learning dan menghitung hasil akurasi, F-Skor, Sensitifi, dan Spesifiti. Hasil tertingginya dalam uji coba mendapatkan nilai akurasi sebesar 82%, F-Skor 83%, Sensitifiti 61%, dan Spesifiti 86% dengan menggunakan fitur fundamental yang diseleksi dan dioversampling.
==================================================================================================================================
Technology is developing very rapidly. One of them is in exchanging videos. Youtube is a popular video exchange website. One category that is often uploaded is music. Music producers must predict the level of success that can be measured using the number of views. This final project predicts the popularity of songs by measuring the number of views and by comparing between Fundamental Features and Technical Features. Fundamental Features is the internal characteristic of the music. Technical Features is the external characteristic of the music. The author uses music videos that are in the "pop music" as a basis for analysis. First built the dataset. Second, extract the fundamental features and technical features from the dataset. Third, preclassification the dataset by stratified cross-validation, normalization, features selection, and oversampling the features. Fourth, do classification using machine learning and calculate the result with accuracy, f-score, sensitivity, and specifity. The highest results in this study is 82% accuracy, 83% precision, 61% sensitivity, and 86% specificity using fundamental features that already selected and oversampled.

Item Type:	Thesis (Other)
Additional Information:	RSIf 006.3 Wij a-1 2019 3100019082222
Uncontrolled Keywords:	Gaussian Naive Bayes, Librosa, Youtube
Subjects:	M Music and Books on Music > M Music M Music and Books on Music > ML Literature of music
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55201-(S1) Undergraduate Thesis
Depositing User:	Unggul Widodo Wijayanto
Date Deposited:	05 Nov 2025 09:46
Last Modified:	05 Nov 2025 09:46
URI:	http://repository.its.ac.id/id/eprint/65571

Actions (login required)

View Item