Comparison of Deep Learning and Machine Learning in Music Genre Categorization - MSc Research Project - Data Analytics

Sep 17, 2026

Uploaded by Theodora

Humans use musical genres to describe and organize diverse music types, and large internet music libraries are often structured via genre hierarchies. Music information retrieval benefits from automatic genre categorization, which can support or replace manual labeling. This study applies machine learning and deep learning models to mp3 tracks from the FMA dataset, extracting features with Mel-spectrograms via librosa and training multiple classifiers. Evaluation shows the CNN model achieves the best performance, with train accuracy of 0.92 and higher F1 scores than other approaches.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages19

Reading Time48 min

AccessFree to download

Abstract

1 Introduction

Comparison of Deep Learning and Machine Learning in Music Genre Categorization MSc Research Project Data Analytics Saviour Nickolas Derel Joseph Fernandez Student ID: 21127051 School of Computing National College of Ireland Supervisor: Mr Vladimir Milosavljevic National College of Ireland Project Submission Sheet School of Computing | Student Name: | Saviour Nickolas Derel Joseph Fernandez | | --- | --- | | Student ID: | 21127051 | | Programme: | Data Analytics | | Year: | 2023 | | Module: | MSc Research Project | | Supervisor: | Mr Vladimir Milosavljevic | | Submission Due Date: | 01/02/2023 | | Project Title: | Comparison of Deep Learning and Machine Learning in Music Genre Categorization | | Word Count: | 4941 | | Page Count: | 17 | I hereby certify that the information contained in this (my submission) is information pertaining to research I conducted for this project. All information other than my own contribution will be fully referenced and listed in the relevant bibliography section at the rear of the project. ALL internet material must be referenced in the bibliography section. Students are required to use the Referencing Standard specified in the report template. To use other author’s written or electronic work is illegal (plagiarism) and may result in disciplinary action. | Signature: | Saviour Nickolas Derel | | --- | --- | | Date: | 30th January 2023 | PLEASE READ THE FOLLOWING INSTRUCTIONS AND CHECKLIST: | Attach a completed copy of this sheet to each project (including multiple copies) . | □ | | --- | --- | | Attach a Moodle submission receipt of the online project submission, to each project (including multiple copies) . | □ | | You must ensure that you retain a HARD COPY of the project, both foryour own reference and in case a project is lost or mislaid. It is not sufficient to keep a copy on computer. | □ | Assignments that are submitted to the Programme Coordinator office must be placed into the assignment box located outside the office. | Office Use Only | | | --- | --- | | Signature: | | | Date: | | | Penalty Applied (if applicable): | | Comparison of Deep Learning and Machine Learning in Music Genre Categorization Saviour Nickolas Derel Joseph Fernandez 21127051 Abstract Humans invented the concept of musical genres to categorize and describe different types of music. The extensive music libraries that are accessible on the internet are frequently organized using genre hierarchies. Music information retrieval systems would benefit greatly from the addition of automatic musical genre categorization, which may supplement or even take the place of the human user in this process where the genres are categorized manually sometimes. There have been variable degrees of success with different music collections, data formats, learning algorithms, and types of neural networks applied. Music indexing and information retrieval both benefit greatly from automatic musical genre classification. The automated classification of musical genres using a combination of machine learning and deep learning algorithms is provided in this study as an effective and efficient approach. The mp3-formatted audio tracks from the FMA dataset are processed for feature extraction using the Mel-spectrogram method for feature retrieval using the ‘librosa’ library. Then the array of converted data from the Mel-spectrogram is used by various classification models. Eight machine-learning includes Na¨ıve Bayes, KNN, Random Forest, XG-Boost classifier, SVM, etc, are used along with the CNN and CNN-LSTM neural network models. On final evaluation, the CNN neural classifier performed better in terms of train model accuracy 0.92 and f1 score than any other classification models and this includes a few parameters tuned for better classification results. 1 Introduction The art of music is the development of aesthetically pleasing or expressive tonal combinations, usually with melody, rhythm, and harmony. We

FAQs

What problem does the study address in music information retrieval?

It addresses how to automatically categorize musical genres to support or supplement manual genre labeling in music libraries.

How are audio features extracted for the classification models?

mp3 tracks from the FMA dataset are processed using Mel-spectrogram feature extraction with the librosa library.

Which model performs best in the final evaluation and what results are reported?

The CNN neural classifier performs best, reporting train model accuracy of 0.92 and superior f1 score compared with other classification models.

Sep 17, 2026

Uploaded by Theodora