aaryanp2904 / spotifysongpopularitypredictorkaggle

Preprocessing and model training code for the Kaggle Spotify Song Popularity Predictor competition.

Spotify Song Popularity Predictor

Directory Structure

• Datasets - Contains the csv files and data values
  • Original - Contains the original test and train .csv files provided
  • Processed - Contains preprocessed data for multiple purposes
    • ModelSelection - This directory has data used for experimentation and finding loss values by dividing the original train.csv file into test and train sets
      • Test - Test data to evaluate models
      • Train - Train data to fit models to
    • FinalDatasets - Contains X and y .csv files for each preprocessing model, the variance and mean of the train data and a standardized and processed version of the original test.csv file
      • PreprocessingModel1 - This is the first preprocessing model and has a different algorithm for accounting for artists
      • PreprocessingModel2 - This is the second preprocessing model and has a different algorithm for accounting for artists
• Models - The .py files that include all the code for the different models
  • Experimenting - These are the models that the ModelSelection/ data was used on to test the accuracy and loss of the models
  • FinalModels - Contains the models and code to generate predictions based on different model architectures
    • NeuralNetworkModelFiles - Contains the .h5 files generated from NeuralNetConstructor.py
• Predictions - Contains the .csv files for each model's prediction, split according to which Preprocessing model's data I used
  • PreprocessingModel1 - Predictions generated using data from the first preprocessing model
  • PreprocessingModel2 - Predictions generated using data from the second preprocessing model
• Preprocessing - Contains the .py files which do all the preprocessing on the original train.csv files, including standardizing numeric features and converting categorical features such as artists and year to suitable numeric equivalents