The goal of this Machine Learning project is to build a model to predict the release year of a song, based on timbre audio features.
Data was collected by the Echo Nest API, which loads music files and returns a JSON file containing much information.
The dataset consists of 515345 instances, 90 predictors and 1 target : Year.
My final model is : KNeighborsClassifier(n_neighbors=3, weights='distance')
It got an accuracy of 5.7 % and an accuracy to the decade of 47.7 %.
Its average absolute distance from the original release year is 9 years.
- A Jupyter Notebook where you can find all data processing, dataviz and machine learning algorithms
- An API
- A PDF file that summarizes my work
- Documentation about the subject, the context and the dataset
- Download the streamlit_app file
- Download the fitted model here : https://drive.google.com/file/d/10WuNKLmeB1-VEW5MpHm7vBf2FiyVpevd/view?usp=sharing
- Or download the dataset here : https://archive.ics.uci.edu/ml/datasets/YearPredictionMSD
- Then download the Jupyter Notebook
- Train the model by yourself
- Save it in a pkl format
- Put the model_final.pkl in streamlit_app/models/
- Install streamlit library :
pip install streamlit - Run the API on Anaconda Prompt or else with this command :
streamlit run ...\streamlit_app\main.py- For example, I enter this command :
streamlit run C:\Users\theop\Desktop\streamlit_app\main.py
- Go to http://localhost:8501
- numpy
- pandas
- pickle