Train in Data conducted a study comparing random forests and gradient boosting models trained on imbalanced data using class weights equal to the imbalance ratio.
Published
Signal category
Research & Knowledge
Quote
“I trained random forests and modern gradient boosting models on 37 benchmark datasets. For each algorithm, I compared models trained on the original class distribution with identical models trained using class weights equal to the imbalance ratio.”
— Soledad Galli|Train in Data team
Company
Train in Data
Online courses on machine learning and data science.
- Industry
- E-Learning Providers
- Location
- Berlin, DE
- Company size
- 3 employees
Our mission is to make intermediate and advanced topics in machine learning, data science, and AI software engineering accessible, understandable, and affordable. Our journey began in 2018, when self-taught data scientist Soledad Galli, tired of spending hours on the internet searching for best practices and information on advanced topics in machine learning, decided to gather this knowledge in easily accessible online courses for the wider data science community. Train in Data now hosts 6 online courses, various books, and an open-source package for feature engineering and selection, Feature-engine, which makes off-the-shelf as well as advanced data transformation techniques accessible to users worldwide.
Founded 2019