Music Genre Classification: A Comparative Analysis of Classical Machine Learning and Deep Learning Approaches

Sachin Prajuli; Abhishek Karna; OmPrakash Dhakl

cs.SDMar 16, 2026

Music Genre Classification: A Comparative Analysis of Classical Machine Learning and Deep Learning Approaches

Authors: Sachin Prajuli, Abhishek Karna, OmPrakash Dhakl

Organizations: Computer Science and Engineering, University of texas at Arlington, Arlington, USA · Electronics and Computer Engineering, Tribhuvan University, IOE, kathmandu · Electronics and Computer Engineering, Tribhuvan University, IOE, Dharan

Abstract

Automatic music genre classification is a long-standing challenge in Music Information Retrieval (MIR); work on non-Western music traditions remains scarce. Nepali music encompasses culturally rich and acoustically diverse genres--from the call-and-response duets of Lok Dohori to the rhythmic poetry of Deuda and the distinctive melodies of Tamang Selo--that have not been addressed by existing classification systems. In this paper, we construct a novel dataset of approximately 8,000 labeled 30-second audio clips spanning eight Nepali music genres and conduct a systematic comparison of nine classification models across two paradigms. Five classical machine learning classifiers (Logistic Regression, SVM, KNN, Random Forest, and XGBoost) are trained on 51 hand-crafted audio features extracted via Librosa, while four deep learning architectures (CNN, RNN, parallel CNN-RNN, and sequential CNN followed by RNN) operate on Mel spectrograms of dimension 640 x 128. Our experiments reveal that the sequential Convolutional Recurrent Neural Network (CRNN)--in which convolutional layers feed into an LSTM--achieves the highest accuracy of 84%, substantially outperforming both the best classical models (Logistic Regression and XGBoost, both at 71%) and all other deep architectures. We provide per-class precision, recall, F1-score, confusion matrices, and ROC analysis for every model, and offer a culturally grounded interpretation of misclassification patterns that reflects genuine overlaps in Nepal's musical traditions.

Figures & tables

Genre	Train	Test
Aadhunik Sangeet	900	100
Deuda	900	100
Tamang Selo	900	100
Lok Dohori	900	100
Purbeli Bhaka	900	100
Rap	900	100

TABLE I : Dataset Composition per Genre

Panel of Figure 2 — Fig. 1 : Mel spectrograms of representative 30-second clips from each genre. Each spectrogram has dimensions $640\times 128$ (time $\times$ Mel bands). Visible differences in energy distribution, harmonic structure, and temporal patterns across genres provide the discriminative cues that the deep learning models learn to exploit.

Model	Accuracy (%)
K-Nearest Neighbors	58
Random Forest	69
Support Vector Machine	69
Logistic Regression	71
XGBoost	71

TABLE II : Classification Accuracy of Classical ML Models (51 Features)

Genre	LR	SVM	KNN	RF	XGBoost
Aadhunik Sangeet	0.68	0.65	0.53	0.64	0.69
Deuda	0.81	0.79	0.67	0.78	0.81
Tamang Selo	0.77	0.71	0.57	0.66	0.73
Lok Dohori	0.64	0.68	0.58	0.61	0.62
Purbeli Bhaka	0.61	0.57	0.37	0.54	0.55
Rap	0.83	0.81	0.74	0.81	0.81

TABLE III : Per-Class F1-Scores for All Classical Machine Learning Models

Architecture	Accuracy (%)
RNN (LSTM only)	68
CNN only	71
Parallel CNN-RNN	76
CRNN (CNN $\rightarrow$ LSTM)	84

TABLE IV : Classification Accuracy of Deep Learning Models (Mel Spectrogram)

Genre	CNN	RNN	Parallel	CRNN (Ours)
Aadhunik Sangeet	0.68	0.68	0.72	0.80
Deuda	0.65	0.73	0.75	0.83
Tamang Selo	0.64	0.78	0.75	0.86
Lok Dohori	0.86	0.67	0.83	0.91
Purbeli Bhaka	0.60	0.51	0.69	0.72
Rap	0.85	0.81	0.86	0.93

TABLE V : Per-Class F1-Scores for All Deep Learning Architectures

Rank	Model	Accuracy (%)
1	CRNN (CNN $\rightarrow$ LSTM)	84
2	Parallel CNN-RNN	76
3	CNN only	71
3	Logistic Regression	71
3	XGBoost	71
6	SVM	69

TABLE VI : Consolidated Accuracy Ranking of All Nine Models

Genre	Prec.	Rec.	F1	Supp.
Aadhunik Sangeet	0.77	0.83	0.80	100
Deuda	0.82	0.83	0.83	100
Tamang Selo	0.88	0.83	0.86	100
Lok Dohori	0.93	0.90	0.91	100
Purbeli Bhaka	0.78	0.67	0.72	100
Rap	0.92	0.93	0.93	100

TABLE VII : Detailed Classification Report for the CRNN Model

Fig. 2 : Confusion matrix for the CRNN model on the test set (800 samples). Diagonal values represent correct classifications. The most frequent confusions occur between Purbeli Bhaka and Deuda, and between Pop and Aadhunik Sangeet—pairs that share genuine cultural and acoustic overlap.

Fig. 3 : Per-class ROC curves for the CRNN model. All eight genres achieve AUC $\geq 0.92$ , with Lok Dohori, Rap, and Rock reaching AUC $\geq 0.99$ .

Panel of Figure 11 — Fig. 4 : Training dynamics for the CRNN model. (a) Training and test accuracy over epochs. (b) Training and test loss over epochs. Test accuracy converges near 84% after $\sim$ 50 epochs.

Explore similar work

YMIR: A new Benchmark Dataset and Model for Arabic Yemeni Music Genre Classification Using Convolutional Neural Networks
Apr 6, 2026Moeen AL-Makhlafi, Abdulrahman A. AlKannad, Eiad Almekhlafi +2Music Information Retrieval Audio Classification
School of Artificial Intelligence, Xidian University, Xi’an, China. · Department of Information Science and Technology, Northwest University, Xi’an 710127, China.
Exploring LLMs for South Asian Music Understanding and Generation
Jun 3, 2026Faria Binte Kader, Mohtasim Hadi Rafi, Shah Wasif Sajjad +1Music Generation Evaluation LLM Evaluation
University of Central Florida · Florida, USA · Auburn University +1
Can pre-trained Deep Learning models predict groove ratings?
Mar 28, 2026Axel Marmoret, Nicolas Farrugia, Jan Alexander StupacherMusic Information Retrieval Audio Representation Learning
IMT Atlantique, Lab-STICC, UMR CNRS 6285, Brest, France · Center for Music In the Brain, Aarhus University, Århus, Denmark

Apr 6, 2026J/K move · Enter open · S save

YMIR: A new Benchmark Dataset and Model for Arabic Yemeni Music Genre Classification Using Convolutional Neural Networks

Moeen AL-Makhlafi, Abdulrahman A. AlKannad, Eiad Almekhlafi, Nawaf Q. Othman Ahmed Mohammed+1

School of Artificial Intelligence, Xidian University, Xi’an, China. · Department of Information Science and Technology, Northwest University, Xi’an 710127, China.

Music Information Retrieval Audio Classification Convolutional Neural Networks

Open paper AskPDF arXiv

Automatic music genre classification is a major task in music information retrieval; however, most current benchmarks and models have been developed primarily for Western music, leaving culturally specific traditions underrepresented. In this paper, we introduce the Yemeni Music Information Retrieval (YMIR) dataset, which contains 1,475 carefully selected audio clips covering five traditional Yemeni genres: Sanaani, Hadhrami, Lahji, Tihami, and Adeni. The dataset was labeled by five Yemeni music experts following a clear and structured protocol, resulting in strong inter-annotator agreement (Fleiss kappa = 0.85). We also propose the Yemeni Music Classification Model (YMCM), a convolutional neural network (CNN)-based system designed to classify music genres from time-frequency features. Using a consistent preprocessing pipeline, we perform a systematic comparison across six experimental groups and five different architectures, resulting in a total of 30 experiments. Specifically, we evaluate several feature representations, including Mel-spectrograms, Chroma, FilterBank, and MFCCs with 13, 20, and 40 coefficients, and benchmark YMCM against standard models (AlexNet, VGG16, MobileNet, and a baseline CNN) under the same experimental conditions. The experimental findings reveal that YMCM is the most effective, achieving the highest accuracy of 98.8% with Mel-spectrogram features. The results also provide practical insights into the relationship between feature representation and model capacity. The findings establish YMIR as a useful benchmark and YMCM as a strong baseline for classifying Yemeni music genres.

Abstract

Explore similar work

YMIR: A new Benchmark Dataset and Model for Arabic Yemeni Music Genre Classification Using Convolutional Neural Networks

Exploring LLMs for South Asian Music Understanding and Generation

Can pre-trained Deep Learning models predict groove ratings?