Add xgboost model #255
Add xgboost model #255
Conversation
|
Check out this pull request on See visual diffs & provide feedback on Jupyter Notebooks. Powered by ReviewNB |
| "metadata": {}, | ||
| "outputs": [], | ||
| "source": [ | ||
| "graph = xgb.to_graphviz(model_basic, num_trees=1, rankdir='LR')\n", |
There was a problem hiding this comment.
this is very nice, but I had to brew install graphviz to make this work (which I think is fine, but might be worth a comment)
There was a problem hiding this comment.
just to note: I also did pip install graphviz before, but that wasn't sufficient it seems (but maybe we should add it to the requirements anyway)
| "source": [ | ||
| "model_ranked, test_data, train_data = xgboost_rank_model(df_xgboost.loc[df_xgboost['points'] > 0], seed=7, test_size=0.33)\n", | ||
| "out = ranked_model_predictions(model_ranked, test_data)" | ||
| ] | ||
| }, | ||
| { | ||
| "cell_type": "code", | ||
| "execution_count": null, | ||
| "metadata": {}, | ||
| "outputs": [], | ||
| "source": [ | ||
| "model_ranked_all, test_data_all, train_data_all = xgboost_rank_model(df_xgboost, seed=7, test_size=0.33)\n", | ||
| "out_all = ranked_model_predictions(model_ranked_all, test_data_all)" | ||
| ] | ||
| }, |
There was a problem hiding this comment.
this is a very good comparison. Interesting that accuracy drops that much when 0's are excluded (though I think we did talk about this at some point). Also, what I don't quite understand is why the accuracy in the basic XGBoost is still 20% higher than in the ranked including 0's.
| "metadata": {}, | ||
| "outputs": [], | ||
| "source": [ | ||
| "violins(out)" |
There was a problem hiding this comment.
those are super helpful to understand what's going on. It seems like there is definitely some predictive quality to the model, though not very precise (except for the top score categories maybe)
Summary
Adds the xgboost model fitting and predictions
What should a reviewer concentrate their feedback on?
Acknowledging contributors