Workshop / Overview

⚠️ A valid COVID certificate must be presented on site to enter the event. ⚠️

In this workshop you will learn how to improve model performance leveraging feature engineering and statistical thinking. We will discuss different approaches to feature engineering and give you a framework that goes beyond the common ""just try out some transformations and see if it works"". You will build a toy forecasting/classification model, and critically assess its results, looking at the model strengths and weaknesses. Together we will use statistical analysis to draw conclusions about the weak points, and use these insights to brainstorm and engineer new features. Eventually you will turn a not-so-ideal model into quite a good one by working with data transformations.

The exercise will be performed in Google Colab.

Workshop / Outcome

Participants will learn how to improve model performance with feature engineering. They will explore how to leverage their statistical knowledge for this task. They will get hands-on experience in how to come up with good ideas for data transformations and assess their impact on the model. 

Workshop / Difficulty

Intermediate level

Workshop / Prerequisites

  • Laptop
  • Intermediate Python knowledge
  • Basic statistics/machine learning knowledge

Track / Co-organizers

Teresa Kubacka

Data Scientist

Magdalena Surówka

Data Scientist, Datali

AMLD EPFL 2021 / Workshops

Towards ethical AI – practical tools for responsible data scientists

With Johan Rochel & Lea Strohm

10:00-11:30 November 10Online

How to make your NLP system multilingual

With Adam Bittlingmayer & Nerses Nersesyan

10:00-12:00 March 02Online

Deep Learning-Driven Text Summarization & Explainability with Reuters News Data

With Nadja Herger, Nina Hristozova & Andreea Iuga

15:00-17:30 March 02Online

AMLD / Global partners