#!/usr/bin/env python
# coding: utf-8

# # Hands-on: Study feature importance in the MAGIC Cherenkov telescope sample

# In[ ]:


import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
from sklearn.model_selection import train_test_split


# In[ ]:


# read data
filename = "https://www.physi.uni-heidelberg.de/~reygers/lectures/2021/ml/data/magic04_data.txt"
df = pd.read_csv(filename, engine='python')

# relabel: gamma shower (g) --> 1 (signal), hadron shower (h) --> 0 (background) 
df['class'] = df['class'].map({'g': 1, 'h': 0})


# In[ ]:


# y = value to predict, X = features
y = df['class'].values
X = df[[col for col in df.columns if col!="class"]]


# In[ ]:


# generate training and test samples
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.1, shuffle=True)


# In[ ]:


# train the normal xgb classifier with all features and calculate the AUC
import xgboost as xgb
from sklearn.metrics import roc_auc_score

XGBclassifier = xgb.sklearn.XGBClassifier(nthread=-1, seed=1, n_estimators=1000)
# YOUR CODE HERE


# #### Now try to train classifiers with $n-1$ features and recalculate the AUC

# In[ ]:


for feat in X.columns:
    # YOUR CODE HERE


# #### How does this compare to the provided `plot_impartance` function from XGBoost?
# (see [XGBoost plotting API](https://xgboost.readthedocs.io/en/latest/python/python_api.html#module-xgboost.plotting)). Do you get the same answer for all three performance measures provided by XGBoost (“weight”, “gain”, or “cover”)?

# In[ ]:


# YOUR CODE HERE