Comment by nwenzel | Hacker News Reader

Comment by nwenzel | Hacker News Reader

Parent Full thread

> Althoff and co used a standard machine learning algorithm to comb through all the possible correlations.

What exactly is a "standard machine learning algorithm"?

I'm sure that probably means that they used something from scikit-learn but "comb[ing] through correlations" isn't as simple as clicking the Go button.

The rest of the article does start to get into labeling, holding out a test set, and some of the data cleanup (the real combing working).

I guess I was just hoping for more detail of how it worked and not that it worked. I get that this wasn't meant to be a PhD thesis on supervised machine learning, but the mechanics of data analysis are really interesting as a process of discovery.

Curious to know what others think. How did the balance of how vs that work for you?

It's a logistic regression model, a basic statistical technique which wouldn't have even come under "machine learning" a few years ago. Later they use some kind of LASSO regression to penalise the inclusion of redundant features.

"Combing through the correlations", it seems, literally means calculating the (Pearson) correlation between two variables (success vs. an input feature) and adding some interpretation, as they do on p6 of the arXiv paper. For test/training data, it looks like they just used a 30/70% split rather than k-fold cross-validation and holdout, but I'm sure it makes no difference either way and in this case (as often) is trivial to design and implement. Presumably their AUROC could be increased just by dropping in an SVM or a Random Forest in place of the logistic regression.

From what I've skimmed of the paper you're over-estimating the complexity of the study.

Great Comment: Having studied graduate statistics within the stats department and a data mining within the CS dept, It amazing how well the CS crowd has rebranded statistics into something you can talk about at a bar without people's eyes glazing over.

Is train/test split part of a normal statistics?

Reply on news.ycombinator.com