first experience with basic java multiclass neural network hosted in azure created for an introductory class on the theme.
given some audio characteristics, classify Spotify songs in their proper class. we will train the algorithm with various songs and their given class. Playlists with pre set themes were used as a starting point, so for example, the songs of a sad playlist, are used to train the algorithm about the what would a "sad" song be like.
all spotify tracks contain some audio characteristics. We will be using:
- acousticness
- danceability
- energy
- instrumentalness
- liveness
- loudness
- tempo
- valence
the meaning of each one of these words can be found in spotify documentation. the dataset used can be found in the main page, and was generated by a python script
- Create A Microsoft Azure Machine Learning Workspace
- Import a dataset
- Create a pipeline and train the model
- Publish the model in a web service
- download the files for the web service
- test the model
- Java code to make the API calls
you need to download this .jar and keep it on the same folder of the Main.java
the java code must be run as:
javac -cp json-simple-1.1.jar Main.java && java -cp .:json-simple-1.1.jar Main