**Goal: Upload a submission that beats the benchmark.**

If you need help getting started, make sure you check out Lesson 4 from 00:45 onward, which covers the Rossman prediction problem - this challenge shares a few similarities.

## Tips

1\. If you use the `ColumnarModelData.get_learner()` learner constructor, be sure to ensure the y_range is between 0 and 1: `md.get_learner(..., y_range=[0, 1])`.

2\. The competition metric uses [ROC AUC](https://developers.google.com/machine-learning/crash-course/classification/roc-and-auc) score. If you want to see the metric while training, you'll need to create a simple custom metric:


    from sklearn.metrics import roc_auc_score
    
    def roc_auc_metric(preds, targs):
        return roc_auc_score(targs.float(), preds.float())
    
    learner.fit(lr, 5, metrics=[roc_auc_metric])


## Dataset Citation

[Moro et al., 2014] S. Moro, P. Cortez and P. Rita. A Data-Driven Approach to Predict the Success of Bank Telemarketing. Decision Support Systems, Elsevier, 62:22-31, June 2014

Thanks to [UCI](https://archive.ics.uci.edu/ml/datasets/bank+marketing)