ESC Score Predictor

Creation of a machine-learning model for predicting scores in the Eurovision Song Contest.

Unfortunately, no useful model was successfully developed. However, the idea and techniques conducted through the project are extensively reported in the Elaborate.pdf.

An example of the model construction is displayed in Modeling.ipynb, but, without the dataset, it cannot be reproduced.

Used Values

Ideally, the following values would be used when calling libROSA's functions for music preprocessing:

import numpy as np

params = {
	'sr': 24576,
	'hop_length': 960,
	'chroma_cqt_hop_length': 3840,
	'tempogram_hop_length': 960,
	'n_fft': 3840,
	'n_mels': 128,
	'n_mfcc': 16,
	'fmin': 0.0,
	'fmax': 11839.82152677230076587824670536366572,
	'chroma_cqt_fmin': 32.70319566257482933473124919041309,
	'frame_length': 1920,
	'kernel_size': 59,
	'win_length': 384,
	'width': 15,
	'dtype': np.float64,
	'stft_dtype': np.complex128
}

However, musicnn's and libROSA's defaults were used in the end.