Predicting Music Track Popularity by Convolutional Neural Networks on Spotify Features and Spectrogram of Audio Waveform

Abstract: In the digital streaming landscape, it's becoming increasingly challenging for artists and industry experts to predict the success of music tracks. This study introduces a pioneering methodology that uses Convolutional Neural Networks (CNNs) and Spotify data analysis to forecast the popularity of music tracks. Our approach takes advantage of Spotify's wide range of features, including acoustic attributes based on the spectrogram of audio waveform, metadata, and user engagement metrics, to capture the complex patterns and relationships that influence a track's popularity. Using a large dataset covering various genres and demographics, our CNN-based model shows impressive effectiveness in predicting the popularity of music tracks. Additionally, we've conducted extensive experiments to assess the strength and adaptability of our model across different musical styles and time periods, with promising results yielding a 97\% F1 score. Our study not only offers valuable insights into the dynamic landscape of digital music consumption but also provides the music industry with advanced predictive tools for assessing and predicting the success of music tracks.

Predicting Music Track Popularity Using CNNs and Spotify Features

The paper presents an innovative approach to predicting music track popularity by leveraging Convolutional Neural Networks (CNNs) with extensive feature data sourced from Spotify. This study underscores the significance of a multifaceted analysis encompassing Spotify's metadata, acoustic attributes derived from the spectrograms of audio waveforms, and user engagement metrics, offering a robust model to anticipate track success. The collaborative effort takes advantage of a large dataset covering various genres and demographics, leading to the development of a CNN-based model that achieves a notable 97% F1 score.

Methodology

The authors utilized Spotify's API to extract a comprehensive dataset comprising approximately 2000 songs from the top 100 artists as of February 2023. The data included a wide range of audio features and metadata specific to the United States. The authors meticulously curated the dataset, ensuring sufficient diversity and reducing data leakage by filtering artists during the training-test split.

Central to the model is the conversion of audio files into spectrograms, with a preference for Mel Spectrograms over traditional Short-Time Fourier Transforms due to their alignment with human auditory perception. This conversion allows the CNN to capture perceptually significant features. The CNN processes both audio-derived features and metadata, creating a concatenated feature vector that encompasses the spectrum of influences on track popularity.

Results

The CNN model demonstrates high effectiveness, achieving a prediction accuracy of 95.68% with a Mean Absolute Error (MAE) of 9.4958. This not only illustrates the model's precision in estimating track popularity but also marks significant improvements over prior methodologies employing various machine learning techniques, such as Random Forests, Support Vector Machines, and deep learning models like HitMusicNet, indicating the advancements made in predictive analytics through sophisticated neural network architectures.

Implications and Future Directions

The findings reveal that artist-related features, namely Artist Popularity and Artist Followers, are more strongly correlated with song popularity than traditional musical attributes. This insight prompts industry stakeholders to focus on enhancing artist profiles as a strategy for boosting track success.

Future work should consider expanding the dataset to further diversify musical attributes and refine the model's adaptability. The exploration of additional data sources and advanced neural processing techniques will further enrich this research avenue. The methodological innovations and remarkable predictive success open exciting pathways for applying machine learning in creative fields, offering both theoretical advancement and practical tools for music industry professionals seeking to strategically plan their releases or analyze market trends.

This study is a decisive step forward in the realm of Hit Song Science, illustrating the transformative role of machine learning in deciphering complex cultural phenomena such as music popularity. It is anticipated that the rapid evolution of digital music consumption will continue to present new opportunities for leveraging data-driven predictive models.