Music Genre Classification
I worked with track-level music data to compare genres through their features, building the preprocessing and training workflow alongside plots of the dataset.
University machine-learning coursework. Data preparation, training and exploratory analysis.
The dataset described ten genres with 5,000 training examples each, and I started by looking at how the available track features were distributed before turning them into inputs for classification, with the exploratory figures kept alongside the preparation code so the transformations could be checked against the data.
The code separates exploratory analysis, preprocessing and training, with plots for features such as energy and danceability and saved validation data and predictions alongside the model workflow.
That separation made it possible to look back at the data transformations and the model output together, though a feature correlation plot describes the dataset and should not be read as a classification result.
The catalogue date follows the first preserved commit on 16 September 2024.
Outcome
A preserved academic classification workflow with exploratory figures and prediction output. The cover shows feature relationships in the dataset rather than an accuracy claim.