random forest bits added

EuroScipy.ipynb

@@ -650,16 +650,15 @@
    },
    "outputs": [],
    "source": [
-    "##### TO BE FIXED!!! ##### \n",
+    "parameters = [{'n_estimators': [5, 10, 20, 50, 100], \n",
+    "               'max_depth': [5, 10, 15]}] \n",
     "\n",
-    "# Define the parameters to be optimised and their values/ranges\n",
-    "\n",
-    "parameters = [{\"n_estimators\": [250, 500, 1000]}]\n",
-    "sample_leaf_options = [1,5,10,50,100,200,500]\n",
-    "\n",
-    "###### WHAT ELSE DO WE ADD HERE?!?!?!!? ###### \n",
+    "grid = GridSearchCV(RandomForestClassifier(), parameters, cv=10)  \n",
+    "grid.fit(XTrain, yTrain)\n",
     "\n",
-    "##### DO WE TUNE WITH A GRID SEARCH??? ##### \n"
+    "best_n_estim = grid.best_params_['n_estimators']\n",
+    "best_max_depth = grid.best_params_['max_depth']\n",
+    "print \"The best parameters are: n_estimators=\", best_n_estim, \" and max_depth=\", best_max_depth"
    ]
   },
   {
@@ -686,7 +685,13 @@
     "####################################################################################  \n",
     "\n",
     "\n",
-    "## WE HAVE NO SOLUTION"
+    "## Solution ## \n",
+    "clfRDF = RandomForestClassifier(n_estimators=best_n_estim, max_depth=best_max_depth)\n",
+    "clfRDF.fit(XTrain, yTrain)\n",
+    "predRF = clfRDF.predict(XTest)\n",
+    "\n",
+    "print metrics.classification_report(yTest, predRF)\n",
+    "print \"Overall Accuracy:\", round(metrics.accuracy_score(yTest, predRF),2)"
    ]
   },
   {
@@ -1356,7 +1361,7 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython2",
-   "version": "2.7.9"
+   "version": "2.7.8"
   }
  },
  "nbformat": 4,