Introduction

This course builds upon Predictive Modeling with Python, covering best practices for building predictive models that perform well in noisy, real-world domains. Feature learning/engineering and model ensembling will be the focus of the course. Examples and exercises will be implemented using Jupyter (formerly iPython) notebooks and the SciKit-Learn library. Participants must have taken Predictive Modeling with Python or obtain instructor permission to enroll.

Who: Anyone who has completed the Predictive Modeling with Python course and wishes to train and deploy predicitve models in real-world domains.

Requirements: Participants must bring a laptop with a few specific software packages installed (see Pre-Workshop Instructions).

Prerequisites: Predictive Modeling with Python short course

Contact: Please mail enalisni@uci.edu for more information.

Tentative Schedule

Syllabus

• Feature Transforms
• Dimensionality Reduction: PCA
• Ensembles of Classifiers (Bagging, Voting, and Stacking)

Pre-Workshop Instructions

See the course’s GitHub page for instructions: https://github.com/UCIDataScienceInitiative/AdvPredictiveModeling_withPython

We’ll expect you to have the Anaconda Python distribution installed: https://www.continuum.io/downloads

Please then run the test notebook to ensure proper installation and to download two datasets: https://github.com/UCIDataScienceInitiative/AdvPredictiveModeling_withPython/blob/master/Test%20Notebook.ipynb

Registration