/kickstarter

Course project for "Data Mining" university course.

Primary LanguageJupyter NotebookMIT LicenseMIT

Kickstarter

Goal

Course project for "Data Mining" university course.

Data

Plan of Attack

  • Load data.
  • Data audit and information value analysis.
  • Feature Engineer: use date columns to make durations (in load data notebook).
  • Finish univariate analysis.
  • PyCaret without text features.
  • Move missing value handling to first notebook.
  • Restructure code.
  • Save PyCaret models plots with performance analysis.
  • Make regressor for number of backers.
  • Add chi-sq.
  • Make regressor for log transform number of backers.
  • Evaluate model.
  • Use regressor and classifier on test set.
  • Experiment with a shallow (non-recurrent) neural network.

Resources