Add Random Forest analysis notebook for Machine Learning 2

M2/Machine Learning 2/Random_forest.Rmd

@@ -0,0 +1,79 @@
+```{r}
+library(randomForest)
+library(MASS)
+library(tree)
+data(Boston)
+```
+
+## Sampling
+
+```{r}
+set.seed(123)
+appr = sample(1:nrow(Boston), nrow(Boston) / 2)
+```
+
+## Regression Tree
+
+```{r}
+arbre.boston = tree(medv ~ ., Boston, subset = appr)
+summary(arbre.boston)
+plot(arbre.boston)
+text(arbre.boston, pretty = 0)
+```
+
+
+```{r}
+cv.boston = cv.tree(arbre.boston)
+plot(cv.boston$size, cv.boston$dev, type = 'b')
+```
+
+
+```{r}
+yprev = predict(arbre.boston, newdata = Boston[-appr, ])
+boston.test = Boston[-appr, "medv"]
+plot(yprev, boston.test)
+abline(0, 1)
+mean((yprev - boston.test)^2)
+```
+
+## Bagging
+
+```{r}
+set.seed(123)
+
+bag.boston = randomForest(
+  medv ~ .,
+  data = Boston,
+  subset = appr,
+  mtry = 13,
+  importance = TRUE
+)
+bag.boston
+yprev.bag = predict(bag.boston, newdata = Boston[-appr, ])
+plot(yprev.bag, boston.test)
+abline(0, 1)
+mean((yprev.bag - boston.test)^2)
+```
+
+## Random Forest
+
+```{r}
+set.seed(123)
+rf.boston = randomForest(
+  medv ~ .,
+  data = Boston,
+  subset = appr,
+  mtry = 6,
+  importance = TRUE
+)
+yprev.rf = predict(rf.boston, newdata = Boston[-appr, ])
+mean((yprev.rf - boston.test)^2)
+```
+
+## Variable importance
+
+
+```{r}
+importance(rf.boston)
+varImpPlot(rf.boston)
+```