Skip to content

A supervised classification model was used to predict a binary variable using 31 contextless features from an anonymised dataset from the Higgs Boson Challenge.

Notifications You must be signed in to change notification settings

irecasens/higgs_boson

Repository files navigation

higgs_boson

A supervised classification model was used to predict a binary variable (described only as ‘class’) using 31 contextless features (‘f1’, ‘f2’, …, ‘f31’). The development was approached in four parts, often overlapping.

First, learners were compared using ‘area under the ROC curve’ (AUC) on a test partition of the data. Second, features (as-given, transformed, and engineered) were selected using forward/backward stepwise search and Top-N. Third, learner parameters were tuned using grid search. Fourth, logistic regression was used as a meta-algorithm to stack and blend models.

The top model uses XGBoost and achieved an AUC of 0.80939 on the Kaggle public leaderboard.

View the report here .

View the code here .

View the original challenge here .

About

A supervised classification model was used to predict a binary variable using 31 contextless features from an anonymised dataset from the Higgs Boson Challenge.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published