diff --git a/jet-tagging.ipynb b/jet-tagging.ipynb
index 318377e..26913ba 100644
--- a/jet-tagging.ipynb
+++ b/jet-tagging.ipynb
@@ -112,7 +112,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 4,
+   "execution_count": null,
    "metadata": {
     "collapsed": false
    },
@@ -132,7 +132,8 @@
     "from sklearn.metrics import roc_auc_score\n",
     "from sklearn.ensemble import GradientBoostingClassifier, GradientBoostingRegressor\n",
     "from sklearn.ensemble import RandomForestClassifier, RandomForestRegressor, ExtraTreesClassifier\n",
-    "from rep.estimators import XGBoostClassifier, SklearnClassifier"
+    "from rep.estimators import XGBoostClassifier, SklearnClassifier\n",
+    "import cPickle as pickle"
    ]
   },
   {
@@ -1021,18 +1022,14 @@
    "cell_type": "code",
    "execution_count": null,
    "metadata": {
-    "collapsed": false
+    "collapsed": true
    },
    "outputs": [],
    "source": [
-    "best_xgb_multi_proba = best_xgb_multi_folding.predict_proba(train_data)"
-   ]
-  },
-  {
-   "cell_type": "markdown",
-   "metadata": {},
-   "source": [
-    "## Results"
+    "# Save classifier\n",
+    "clf_pickle = open('best_xgb_multi_folding.pkl', 'wb')\n",
+    "#pickle.dump(best_xgb_multi_folding, clf_pickle)\n",
+    "clf_pickle.close()"
    ]
   },
   {
@@ -1043,260 +1040,21 @@
    },
    "outputs": [],
    "source": [
-    "best_xgb_multi_results = get_result(best_xgb_multi_proba, labels, \"best_xgb_multiclass\")"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": 200,
-   "metadata": {
-    "collapsed": false
-   },
-   "outputs": [
-    {
-     "data": {
-      "text/html": [
-       "<div>\n",
-       "<table border=\"1\" class=\"dataframe\">\n",
-       "  <thead>\n",
-       "    <tr style=\"text-align: right;\">\n",
-       "      <th></th>\n",
-       "      <th>b vs c</th>\n",
-       "      <th>b vs light</th>\n",
-       "      <th>c vs light</th>\n",
-       "    </tr>\n",
-       "    <tr>\n",
-       "      <th>name</th>\n",
-       "      <th></th>\n",
-       "      <th></th>\n",
-       "      <th></th>\n",
-       "    </tr>\n",
-       "  </thead>\n",
-       "  <tbody>\n",
-       "    <tr>\n",
-       "      <th>multiclass</th>\n",
-       "      <td>0.949536</td>\n",
-       "      <td>0.983856</td>\n",
-       "      <td>0.973569</td>\n",
-       "    </tr>\n",
-       "    <tr>\n",
-       "      <th>pca_multiclass</th>\n",
-       "      <td>0.945119</td>\n",
-       "      <td>0.985436</td>\n",
-       "      <td>0.977200</td>\n",
-       "    </tr>\n",
-       "    <tr>\n",
-       "      <th>sel_new_multiclass</th>\n",
-       "      <td>0.950772</td>\n",
-       "      <td>0.988023</td>\n",
-       "      <td>0.980581</td>\n",
-       "    </tr>\n",
-       "    <tr>\n",
-       "      <th>best_xgb_multiclass</th>\n",
-       "      <td>0.949341</td>\n",
-       "      <td>0.984908</td>\n",
-       "      <td>0.975656</td>\n",
-       "    </tr>\n",
-       "  </tbody>\n",
-       "</table>\n",
-       "</div>"
-      ],
-      "text/plain": [
-       "                       b vs c  b vs light  c vs light\n",
-       "name                                                 \n",
-       "multiclass           0.949536    0.983856    0.973569\n",
-       "pca_multiclass       0.945119    0.985436    0.977200\n",
-       "sel_new_multiclass   0.950772    0.988023    0.980581\n",
-       "best_xgb_multiclass  0.949341    0.984908    0.975656"
-      ]
-     },
-     "execution_count": 200,
-     "metadata": {},
-     "output_type": "execute_result"
-    }
-   ],
-   "source": [
-    "result = pandas.concat([multiclass_result, \n",
-    "                        pca_multiclass_result,  \n",
-    "                        sel_new_multi_results,\n",
-    "                        best_xgb_multi_results])\n",
-    "result.index = result['name']\n",
-    "result = result.drop('name', axis=1)\n",
-    "result"
-   ]
-  },
-  {
-   "cell_type": "markdown",
-   "metadata": {},
-   "source": [
-    "# Grid search RF"
-   ]
-  },
-  {
-   "cell_type": "markdown",
-   "metadata": {},
-   "source": [
-    "## Search"
+    "# Read classifier\n",
+    "clf_pickle2 = open('best_xgb_multi_folding.pkl', 'rb')\n",
+    "#best_xgb_multi_folding = pickle.load(clf_pickle2)\n",
+    "clf_pickle2.close()"
    ]
   },
   {
    "cell_type": "code",
-   "execution_count": 215,
-   "metadata": {
-    "collapsed": false
-   },
-   "outputs": [
-    {
-     "data": {
-      "text/plain": [
-       "GridSearchCV(cv=2, error_score='raise',\n",
-       "       estimator=RandomForestClassifier(bootstrap=True, class_weight='auto', criterion='gini',\n",
-       "            max_depth=None, max_features='auto', max_leaf_nodes=None,\n",
-       "            min_samples_leaf=1, min_samples_split=2,\n",
-       "            min_weight_fraction_leaf=0.0, n_estimators=1000, n_jobs=1,\n",
-       "            oob_score=False, random_state=None, verbose=0,\n",
-       "            warm_start=False),\n",
-       "       fit_params={}, iid=True, loss_func=None, n_jobs=8,\n",
-       "       param_grid={'max_features': ['auto', 0.3, 0.5, 1.0], 'max_depth': [4, 8, 12, None]},\n",
-       "       pre_dispatch='2*n_jobs', refit=True, score_func=None,\n",
-       "       scoring=<__main__.MyMeanRocAucScorer object at 0x7f14ac0ec4d0>,\n",
-       "       verbose=0)"
-      ]
-     },
-     "execution_count": 215,
-     "metadata": {},
-     "output_type": "execute_result"
-    }
-   ],
-   "source": [
-    "from sklearn.grid_search import GridSearchCV\n",
-    "from rep.utils import train_test_split\n",
-    "\n",
-    "train_train_data, test_data, train_labels, test_labels = train_test_split(train_data, labels, test_size=0.3)\n",
-    "\n",
-    "grid_param = {}\n",
-    "grid_param['max_features'] = ['auto', 0.3, 0.5, 1.0]\n",
-    "grid_param['max_depth'] = [4, 8, 12, None]\n",
-    "\n",
-    "rf_base = RandomForestClassifier(n_estimators=1000, max_depth=None, max_features='auto', class_weight='auto')\n",
-    "\n",
-    "grid_rf = GridSearchCV(estimator=rf_base, param_grid=grid_param, cv=2, \n",
-    "                    scoring=MyMeanRocAucScorer(test_data, test_labels), n_jobs=8)\n",
-    "grid_rf.fit(train_train_data, train_labels)"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": 216,
-   "metadata": {
-    "collapsed": false
-   },
-   "outputs": [
-    {
-     "data": {
-      "text/plain": [
-       "{'max_depth': 12, 'max_features': 'auto'}"
-      ]
-     },
-     "execution_count": 216,
-     "metadata": {},
-     "output_type": "execute_result"
-    }
-   ],
-   "source": [
-    "grid_rf.best_params_"
-   ]
-  },
-  {
-   "cell_type": "markdown",
-   "metadata": {},
-   "source": [
-    "## Train"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": 217,
-   "metadata": {
-    "collapsed": false
-   },
-   "outputs": [
-    {
-     "name": "stdout",
-     "output_type": "stream",
-     "text": [
-      "CPU times: user 2min 16s, sys: 216 ms, total: 2min 16s\n",
-      "Wall time: 2min 16s\n"
-     ]
-    },
-    {
-     "data": {
-      "text/plain": [
-       "FoldingClassifier(base_estimator=RandomForestClassifier(bootstrap=True, class_weight='auto', criterion='gini',\n",
-       "            max_depth=12, max_features='auto', max_leaf_nodes=None,\n",
-       "            min_samples_leaf=1, min_samples_split=2,\n",
-       "            min_weight_fraction_leaf=0.0, n_estimators=1000, n_jobs=1,\n",
-       "            oob_score=False, random_state=None, verbose=0,\n",
-       "            warm_start=False),\n",
-       "         features=['Feature_0', 'Feature_1', 'Feature_2', 'Feature_3', 'Feature_4', 'Feature_5', 'Feature_6', 'Feature_7', 'Feature_8'],\n",
-       "         n_folds=2, parallel_profile=None, random_state=11)"
-      ]
-     },
-     "execution_count": 217,
-     "metadata": {},
-     "output_type": "execute_result"
-    }
-   ],
-   "source": [
-    "rf_best = grid_rf.best_estimator_\n",
-    "best_rf_multi_folding = FoldingClassifier(rf_best, n_folds=2, random_state=11)\n",
-    "%time best_rf_multi_folding.fit(train_data, labels)"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": 218,
-   "metadata": {
-    "collapsed": false
-   },
-   "outputs": [
-    {
-     "data": {
-      "text/plain": [
-       "<bound method RandomForestClassifier.get_params of RandomForestClassifier(bootstrap=True, class_weight='auto', criterion='gini',\n",
-       "            max_depth=12, max_features='auto', max_leaf_nodes=None,\n",
-       "            min_samples_leaf=1, min_samples_split=2,\n",
-       "            min_weight_fraction_leaf=0.0, n_estimators=1000, n_jobs=1,\n",
-       "            oob_score=False, random_state=None, verbose=0,\n",
-       "            warm_start=False)>"
-      ]
-     },
-     "execution_count": 218,
-     "metadata": {},
-     "output_type": "execute_result"
-    }
-   ],
-   "source": [
-    "rf_best.get_params"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": 219,
+   "execution_count": null,
    "metadata": {
     "collapsed": false
    },
-   "outputs": [
-    {
-     "name": "stdout",
-     "output_type": "stream",
-     "text": [
-      "KFold prediction using folds column\n"
-     ]
-    }
-   ],
+   "outputs": [],
    "source": [
-    "best_rf_multi_proba = best_rf_multi_folding.predict_proba(train_data)"
+    "best_xgb_multi_proba = best_xgb_multi_folding.predict_proba(train_data)"
    ]
   },
   {
@@ -1308,18 +1066,18 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 220,
+   "execution_count": null,
    "metadata": {
     "collapsed": true
    },
    "outputs": [],
    "source": [
-    "best_rf_multi_results = get_result(best_rf_multi_proba, labels, \"best_rf_multiclass\")"
+    "best_xgb_multi_results = get_result(best_xgb_multi_proba, labels, \"best_xgb_multiclass\")"
    ]
   },
   {
    "cell_type": "code",
-   "execution_count": 221,
+   "execution_count": 200,
    "metadata": {
     "collapsed": false
    },
@@ -1368,12 +1126,6 @@
        "      <td>0.984908</td>\n",
        "      <td>0.975656</td>\n",
        "    </tr>\n",
-       "    <tr>\n",
-       "      <th>best_rf_multiclass</th>\n",
-       "      <td>0.947559</td>\n",
-       "      <td>0.981816</td>\n",
-       "      <td>0.968406</td>\n",
-       "    </tr>\n",
        "  </tbody>\n",
        "</table>\n",
        "</div>"
@@ -1384,11 +1136,10 @@
        "multiclass           0.949536    0.983856    0.973569\n",
        "pca_multiclass       0.945119    0.985436    0.977200\n",
        "sel_new_multiclass   0.950772    0.988023    0.980581\n",
-       "best_xgb_multiclass  0.949341    0.984908    0.975656\n",
-       "best_rf_multiclass   0.947559    0.981816    0.968406"
+       "best_xgb_multiclass  0.949341    0.984908    0.975656"
       ]
      },
-     "execution_count": 221,
+     "execution_count": 200,
      "metadata": {},
      "output_type": "execute_result"
     }
@@ -1397,8 +1148,7 @@
     "result = pandas.concat([multiclass_result, \n",
     "                        pca_multiclass_result,  \n",
     "                        sel_new_multi_results,\n",
-    "                        best_xgb_multi_results,\n",
-    "                        best_rf_multi_results])\n",
+    "                        best_xgb_multi_results])\n",
     "result.index = result['name']\n",
     "result = result.drop('name', axis=1)\n",
     "result"
@@ -1408,7 +1158,7 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "# Grid search AdaBoost"
+    "# Grid search XGBoost + features generation + features selection"
    ]
   },
   {
@@ -1422,32 +1172,33 @@
    "cell_type": "code",
    "execution_count": null,
    "metadata": {
-    "collapsed": false
+    "collapsed": true
    },
    "outputs": [],
    "source": [
     "from sklearn.grid_search import GridSearchCV\n",
     "from rep.utils import train_test_split\n",
-    "from sklearn.ensemble import AdaBoostClassifier\n",
-    "from sklearn.tree import DecisionTreeClassifier\n",
     "\n",
-    "train_train_data, test_data, train_labels, test_labels = train_test_split(train_data, labels, test_size=0.3)\n",
+    "sel_new_train_train_data, sel_new_test_data, sel_new_train_labels, sel_new_test_labels = \\\n",
+    "train_test_split(sel_new_train_data, labels, test_size=0.3)\n",
     "\n",
     "grid_param = {}\n",
-    "grid_param['learning_rate'] = [0.1, 0.05, 0.01]\n",
-    "grid_param['max_depth'] = [4, 8, 12, None]\n",
+    "grid_param['eta'] = [0.1, 0.05, 0.01]\n",
+    "grid_param['max_depth'] = [6, 8, 12, 100]\n",
+    "grid_param['colsample'] = [0.3, 0.5, 0.7, 1.0]\n",
+    "grid_param['subsample'] = [0.3, 0.5, 0.7, 1.0]\n",
     "\n",
-    "ada_base = AdaBoostClassifier(base_estimator=DecisionTreeClassifier(max_depth=8, max_features='auto', class_weight='balanced'),\n",
-    "                             n_estimators=1000, learning_rate=0.01)\n",
+    "xgb_base = XGBoostClassifier(n_estimators=1000, colsample=0.7, eta=0.01, nthreads=1, \n",
+    "                             subsample=0.5, max_depth=8)\n",
     "\n",
-    "grid_ada = GridSearchCV(estimator=ada_base, param_grid=grid_param, cv=2, \n",
-    "                    scoring=MyMeanRocAucScorer(test_data, test_labels), n_jobs=8)\n",
-    "grid_ada.fit(train_train_data, train_labels)"
+    "sel_new_grid_xgb = GridSearchCV(estimator=xgb_base, param_grid=grid_param, cv=2, \n",
+    "                    scoring=MyMeanRocAucScorer(sel_new_test_data, sel_new_test_labels), n_jobs=8)\n",
+    "sel_new_grid_xgb.fit(sel_new_train_train_data, sel_new_train_labels)"
    ]
   },
   {
    "cell_type": "code",
-   "execution_count": 212,
+   "execution_count": 241,
    "metadata": {
     "collapsed": false
    },
@@ -1455,16 +1206,16 @@
     {
      "data": {
       "text/plain": [
-       "{'learning_rate': 0.1}"
+       "{'colsample': 0.7, 'eta': 0.01, 'max_depth': 6, 'subsample': 0.5}"
       ]
      },
-     "execution_count": 212,
+     "execution_count": 241,
      "metadata": {},
      "output_type": "execute_result"
     }
    ],
    "source": [
-    "grid_ada.best_params_"
+    "sel_new_grid_xgb.best_params_"
    ]
   },
   {
@@ -1474,37 +1225,6 @@
     "## Train"
    ]
   },
-  {
-   "cell_type": "code",
-   "execution_count": null,
-   "metadata": {
-    "collapsed": false
-   },
-   "outputs": [],
-   "source": [
-    "ada_best = grid_ada.best_estimator_\n",
-    "best_ada_multi_folding = FoldingClassifier(ada_best, n_folds=2, random_state=11)\n",
-    "%time best_ada_multi_folding.fit(train_data, labels)"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": null,
-   "metadata": {
-    "collapsed": false
-   },
-   "outputs": [],
-   "source": [
-    "best_ada_multi_proba = best_ada_multi_folding.predict_proba(train_data)"
-   ]
-  },
-  {
-   "cell_type": "markdown",
-   "metadata": {},
-   "source": [
-    "## Results"
-   ]
-  },
   {
    "cell_type": "code",
    "execution_count": null,
@@ -1513,193 +1233,48 @@
    },
    "outputs": [],
    "source": [
-    "best_ada_multi_results = get_result(best_ada_multi_proba, labels, \"best_ada_multiclass\")"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": 222,
-   "metadata": {
-    "collapsed": false
-   },
-   "outputs": [
-    {
-     "data": {
-      "text/html": [
-       "<div>\n",
-       "<table border=\"1\" class=\"dataframe\">\n",
-       "  <thead>\n",
-       "    <tr style=\"text-align: right;\">\n",
-       "      <th></th>\n",
-       "      <th>b vs c</th>\n",
-       "      <th>b vs light</th>\n",
-       "      <th>c vs light</th>\n",
-       "    </tr>\n",
-       "    <tr>\n",
-       "      <th>name</th>\n",
-       "      <th></th>\n",
-       "      <th></th>\n",
-       "      <th></th>\n",
-       "    </tr>\n",
-       "  </thead>\n",
-       "  <tbody>\n",
-       "    <tr>\n",
-       "      <th>multiclass</th>\n",
-       "      <td>0.949536</td>\n",
-       "      <td>0.983856</td>\n",
-       "      <td>0.973569</td>\n",
-       "    </tr>\n",
-       "    <tr>\n",
-       "      <th>pca_multiclass</th>\n",
-       "      <td>0.945119</td>\n",
-       "      <td>0.985436</td>\n",
-       "      <td>0.977200</td>\n",
-       "    </tr>\n",
-       "    <tr>\n",
-       "      <th>sel_new_multiclass</th>\n",
-       "      <td>0.950772</td>\n",
-       "      <td>0.988023</td>\n",
-       "      <td>0.980581</td>\n",
-       "    </tr>\n",
-       "    <tr>\n",
-       "      <th>best_xgb_multiclass</th>\n",
-       "      <td>0.949341</td>\n",
-       "      <td>0.984908</td>\n",
-       "      <td>0.975656</td>\n",
-       "    </tr>\n",
-       "    <tr>\n",
-       "      <th>best_rf_multiclass</th>\n",
-       "      <td>0.947559</td>\n",
-       "      <td>0.981816</td>\n",
-       "      <td>0.968406</td>\n",
-       "    </tr>\n",
-       "    <tr>\n",
-       "      <th>best_ada_multiclass</th>\n",
-       "      <td>0.939101</td>\n",
-       "      <td>0.977499</td>\n",
-       "      <td>0.958886</td>\n",
-       "    </tr>\n",
-       "  </tbody>\n",
-       "</table>\n",
-       "</div>"
-      ],
-      "text/plain": [
-       "                       b vs c  b vs light  c vs light\n",
-       "name                                                 \n",
-       "multiclass           0.949536    0.983856    0.973569\n",
-       "pca_multiclass       0.945119    0.985436    0.977200\n",
-       "sel_new_multiclass   0.950772    0.988023    0.980581\n",
-       "best_xgb_multiclass  0.949341    0.984908    0.975656\n",
-       "best_rf_multiclass   0.947559    0.981816    0.968406\n",
-       "best_ada_multiclass  0.939101    0.977499    0.958886"
-      ]
-     },
-     "execution_count": 222,
-     "metadata": {},
-     "output_type": "execute_result"
-    }
-   ],
-   "source": [
-    "result = pandas.concat([multiclass_result, \n",
-    "                        pca_multiclass_result,  \n",
-    "                        sel_new_multi_results,\n",
-    "                        best_xgb_multi_results,\n",
-    "                        best_rf_multi_results,\n",
-    "                        best_ada_multi_results])\n",
-    "result.index = result['name']\n",
-    "result = result.drop('name', axis=1)\n",
-    "result"
-   ]
-  },
-  {
-   "cell_type": "markdown",
-   "metadata": {},
-   "source": [
-    "# Grid search ExtraTreeClassifier"
-   ]
-  },
-  {
-   "cell_type": "markdown",
-   "metadata": {},
-   "source": [
-    "## Search"
+    "sel_new_xgb_best = sel_new_grid_xgb.best_estimator_.set_params(features=None)\n",
+    "sel_new_best_xgb_multi_folding = FoldingClassifier(sel_new_xgb_best, n_folds=2, random_state=11)\n",
+    "%time sel_new_best_xgb_multi_folding.fit(sel_new_train_data, labels)"
    ]
   },
   {
    "cell_type": "code",
    "execution_count": null,
    "metadata": {
-    "collapsed": false
+    "collapsed": true
    },
    "outputs": [],
    "source": [
-    "from sklearn.grid_search import GridSearchCV\n",
-    "from rep.utils import train_test_split\n",
-    "\n",
-    "train_train_data, test_data, train_labels, test_labels = train_test_split(train_data, labels, test_size=0.3)\n",
-    "\n",
-    "grid_param = {}\n",
-    "grid_param['max_features'] = ['auto', 0.3, 0.5, 0.7, 1.0]\n",
-    "grid_param['max_depth'] = [4, 8, 12, None]\n",
-    "\n",
-    "ex_base = ExtraTreesClassifier(n_estimators=1000, max_depth=6, max_features='auto', class_weight='balanced', n_jobs=4)\n",
-    "\n",
-    "ex_base = GridSearchCV(estimator=ex_base, param_grid=grid_param, cv=2, \n",
-    "                    scoring=MyMeanRocAucScorer(test_data, test_labels), n_jobs=8)\n",
-    "ex_base.fit(train_train_data, train_labels)"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": 213,
-   "metadata": {
-    "collapsed": false
-   },
-   "outputs": [
-    {
-     "data": {
-      "text/plain": [
-       "{'max_depth': 8, 'max_features': 1.0}"
-      ]
-     },
-     "execution_count": 213,
-     "metadata": {},
-     "output_type": "execute_result"
-    }
-   ],
-   "source": [
-    "ex_base.best_params_"
-   ]
-  },
-  {
-   "cell_type": "markdown",
-   "metadata": {},
-   "source": [
-    "## Train"
+    "# Save classifier\n",
+    "clf_pickle = open('sel_new_best_xgb_multi_folding.pkl', 'wb')\n",
+    "#pickle.dump(sel_new_best_xgb_multi_folding, clf_pickle)\n",
+    "clf_pickle.close()"
    ]
   },
   {
    "cell_type": "code",
    "execution_count": null,
    "metadata": {
-    "collapsed": false
+    "collapsed": true
    },
    "outputs": [],
    "source": [
-    "ex_best = ex_base.best_estimator_\n",
-    "best_ex_multi_folding = FoldingClassifier(ex_best, n_folds=2, random_state=11)\n",
-    "%time best_ex_multi_folding.fit(train_data, labels)"
+    "# Read classifier\n",
+    "clf_pickle2 = open('sel_new_best_xgb_multi_folding.pkl', 'rb')\n",
+    "#sel_new_best_xgb_multi_folding = pickle.load(clf_pickle2)\n",
+    "clf_pickle2.close()"
    ]
   },
   {
    "cell_type": "code",
    "execution_count": null,
    "metadata": {
-    "collapsed": false
+    "collapsed": true
    },
    "outputs": [],
    "source": [
-    "best_ex_multi_proba = best_ex_multi_folding.predict_proba(train_data)"
+    "sel_new_best_xgb_multi_proba = sel_new_best_xgb_multi_folding.predict_proba(sel_new_train_data)"
    ]
   },
   {
@@ -1717,12 +1292,12 @@
    },
    "outputs": [],
    "source": [
-    "best_ex_multi_results = get_result(best_ex_multi_proba, labels, \"best_ex_multiclass\")"
+    "sel_new_best_xgb_multi_results = get_result(sel_new_best_xgb_multi_proba, labels, \"sel_new_best_xgb_multiclass\")"
    ]
   },
   {
    "cell_type": "code",
-   "execution_count": 223,
+   "execution_count": 242,
    "metadata": {
     "collapsed": false
    },
@@ -1772,181 +1347,26 @@
        "      <td>0.975656</td>\n",
        "    </tr>\n",
        "    <tr>\n",
-       "      <th>best_rf_multiclass</th>\n",
-       "      <td>0.947559</td>\n",
-       "      <td>0.981816</td>\n",
-       "      <td>0.968406</td>\n",
-       "    </tr>\n",
-       "    <tr>\n",
-       "      <th>best_ada_multiclass</th>\n",
-       "      <td>0.939101</td>\n",
-       "      <td>0.977499</td>\n",
-       "      <td>0.958886</td>\n",
-       "    </tr>\n",
-       "    <tr>\n",
-       "      <th>best_ex_multiclass</th>\n",
-       "      <td>0.934843</td>\n",
-       "      <td>0.973519</td>\n",
-       "      <td>0.953799</td>\n",
+       "      <th>sel_new_best_xgb_multiclass</th>\n",
+       "      <td>0.951279</td>\n",
+       "      <td>0.987887</td>\n",
+       "      <td>0.980127</td>\n",
        "    </tr>\n",
        "  </tbody>\n",
        "</table>\n",
        "</div>"
       ],
       "text/plain": [
-       "                       b vs c  b vs light  c vs light\n",
-       "name                                                 \n",
-       "multiclass           0.949536    0.983856    0.973569\n",
-       "pca_multiclass       0.945119    0.985436    0.977200\n",
-       "sel_new_multiclass   0.950772    0.988023    0.980581\n",
-       "best_xgb_multiclass  0.949341    0.984908    0.975656\n",
-       "best_rf_multiclass   0.947559    0.981816    0.968406\n",
-       "best_ada_multiclass  0.939101    0.977499    0.958886\n",
-       "best_ex_multiclass   0.934843    0.973519    0.953799"
-      ]
-     },
-     "execution_count": 223,
-     "metadata": {},
-     "output_type": "execute_result"
-    }
-   ],
-   "source": [
-    "result = pandas.concat([multiclass_result, \n",
-    "                        pca_multiclass_result,  \n",
-    "                        sel_new_multi_results,\n",
-    "                        best_xgb_multi_results,\n",
-    "                        best_rf_multi_results,\n",
-    "                        best_ada_multi_results,\n",
-    "                        best_ex_multi_results])\n",
-    "result.index = result['name']\n",
-    "result = result.drop('name', axis=1)\n",
-    "result"
-   ]
-  },
-  {
-   "cell_type": "markdown",
-   "metadata": {
-    "collapsed": true
-   },
-   "source": [
-    "# Combination"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": 230,
-   "metadata": {
-    "collapsed": true
-   },
-   "outputs": [],
-   "source": [
-    "comb_proba = 0.5 * (best_xgb_multi_proba + best_rf_multi_proba)"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": 231,
-   "metadata": {
-    "collapsed": true
-   },
-   "outputs": [],
-   "source": [
-    "comb_multi_results = get_result(comb_proba, labels, \"comb_multiclass\")"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": 232,
-   "metadata": {
-    "collapsed": false
-   },
-   "outputs": [
-    {
-     "data": {
-      "text/html": [
-       "<div>\n",
-       "<table border=\"1\" class=\"dataframe\">\n",
-       "  <thead>\n",
-       "    <tr style=\"text-align: right;\">\n",
-       "      <th></th>\n",
-       "      <th>b vs c</th>\n",
-       "      <th>b vs light</th>\n",
-       "      <th>c vs light</th>\n",
-       "    </tr>\n",
-       "    <tr>\n",
-       "      <th>name</th>\n",
-       "      <th></th>\n",
-       "      <th></th>\n",
-       "      <th></th>\n",
-       "    </tr>\n",
-       "  </thead>\n",
-       "  <tbody>\n",
-       "    <tr>\n",
-       "      <th>multiclass</th>\n",
-       "      <td>0.949536</td>\n",
-       "      <td>0.983856</td>\n",
-       "      <td>0.973569</td>\n",
-       "    </tr>\n",
-       "    <tr>\n",
-       "      <th>pca_multiclass</th>\n",
-       "      <td>0.945119</td>\n",
-       "      <td>0.985436</td>\n",
-       "      <td>0.977200</td>\n",
-       "    </tr>\n",
-       "    <tr>\n",
-       "      <th>sel_new_multiclass</th>\n",
-       "      <td>0.950772</td>\n",
-       "      <td>0.988023</td>\n",
-       "      <td>0.980581</td>\n",
-       "    </tr>\n",
-       "    <tr>\n",
-       "      <th>best_xgb_multiclass</th>\n",
-       "      <td>0.949341</td>\n",
-       "      <td>0.984908</td>\n",
-       "      <td>0.975656</td>\n",
-       "    </tr>\n",
-       "    <tr>\n",
-       "      <th>best_rf_multiclass</th>\n",
-       "      <td>0.947559</td>\n",
-       "      <td>0.981816</td>\n",
-       "      <td>0.968406</td>\n",
-       "    </tr>\n",
-       "    <tr>\n",
-       "      <th>best_ada_multiclass</th>\n",
-       "      <td>0.939101</td>\n",
-       "      <td>0.977499</td>\n",
-       "      <td>0.958886</td>\n",
-       "    </tr>\n",
-       "    <tr>\n",
-       "      <th>best_ex_multiclass</th>\n",
-       "      <td>0.934843</td>\n",
-       "      <td>0.973519</td>\n",
-       "      <td>0.953799</td>\n",
-       "    </tr>\n",
-       "    <tr>\n",
-       "      <th>comb_multiclass</th>\n",
-       "      <td>0.949307</td>\n",
-       "      <td>0.983889</td>\n",
-       "      <td>0.973104</td>\n",
-       "    </tr>\n",
-       "  </tbody>\n",
-       "</table>\n",
-       "</div>"
-      ],
-      "text/plain": [
-       "                       b vs c  b vs light  c vs light\n",
-       "name                                                 \n",
-       "multiclass           0.949536    0.983856    0.973569\n",
-       "pca_multiclass       0.945119    0.985436    0.977200\n",
-       "sel_new_multiclass   0.950772    0.988023    0.980581\n",
-       "best_xgb_multiclass  0.949341    0.984908    0.975656\n",
-       "best_rf_multiclass   0.947559    0.981816    0.968406\n",
-       "best_ada_multiclass  0.939101    0.977499    0.958886\n",
-       "best_ex_multiclass   0.934843    0.973519    0.953799\n",
-       "comb_multiclass      0.949307    0.983889    0.973104"
+       "                               b vs c  b vs light  c vs light\n",
+       "name                                                         \n",
+       "multiclass                   0.949536    0.983856    0.973569\n",
+       "pca_multiclass               0.945119    0.985436    0.977200\n",
+       "sel_new_multiclass           0.950772    0.988023    0.980581\n",
+       "best_xgb_multiclass          0.949341    0.984908    0.975656\n",
+       "sel_new_best_xgb_multiclass  0.951279    0.987887    0.980127"
       ]
      },
-     "execution_count": 232,
+     "execution_count": 242,
      "metadata": {},
      "output_type": "execute_result"
     }
@@ -1956,10 +1376,7 @@
     "                        pca_multiclass_result,  \n",
     "                        sel_new_multi_results,\n",
     "                        best_xgb_multi_results,\n",
-    "                        best_rf_multi_results,\n",
-    "                        best_ada_multi_results,\n",
-    "                        best_ex_multi_results,\n",
-    "                        comb_multi_results])\n",
+    "                        sel_new_best_xgb_multi_results])\n",
     "result.index = result['name']\n",
     "result = result.drop('name', axis=1)\n",
     "result"