Skip to content

Tuna – a simple streaming ETL for machine learning features

github.com/MaxHalford
1 pointLemaxoxo1 comment
On HN

Comments

Hello everyone,

I'm currently participating in the "PLAsTiCC Astronomical Classification" Kaggle competition. The dataset is rather large (~5M rows) so I decided to write a simple tool to compute aggregate features online. I got really inspired and tidied things up over the weekend. Maybe this can be of interest to some of you.

PS: I'm aware that there are other similar projects out there such as Spark Streaming, but they all feel too bloated and difficult to grok.

AboutSource Built by g1lg1l

Hackerly is an independent reader for Hacker News, built on the public HN API. Not affiliated with Y Combinator.