--- title: "Regression" output: pdf_document: fig_height: 3 --- # Data Set -- Boston Housing ```{r} library(mlr) # mlr comes with example tasks bh.task ``` ```{r} # this is what it looks like... head(getTaskData(bh.task)) ``` ```{r} # ...and this is what it looks like plotted library(ggplot2) ggplot(getTaskData(bh.task), aes(lstat, rm)) + geom_point(aes(color = medv)) ``` # Linear Model ```{r} learner = makeLearner("regr.lm") rdesc = makeResampleDesc(method = "Holdout", split = 2/3) result = resample(learner, bh.task, rdesc, models = TRUE) getRRPredictions(result) ``` ```{r} getLearnerModel(result$models[[1]]) ``` ```{r} plot(getLearnerModel(result$models[[1]])) ``` ```{r} plotLearnerPrediction(learner, bh.task, features = c("lstat")) ``` ```{r} plotLearnerPrediction(learner, bh.task, features = c("lstat", "rm")) ``` # Regression Splines ```{r} learner = makeLearner("regr.earth") result = resample(learner, bh.task, rdesc, models = TRUE) getRRPredictions(result) ``` ```{r} getLearnerModel(result$models[[1]]) ``` ```{r} plot(getLearnerModel(result$models[[1]])) ``` ```{r} plotLearnerPrediction(learner, bh.task, features = c("lstat")) ``` ```{r} plotLearnerPrediction(learner, bh.task, features = c("lstat", "rm")) ``` # Boosting ```{r} learner = makeLearner("regr.blackboost") result = resample(learner, bh.task, rdesc, models = TRUE) getRRPredictions(result) ``` ```{r} getLearnerModel(result$models[[1]]) ``` ```{r} plotLearnerPrediction(learner, bh.task, features = c("lstat")) ``` ```{r} plotLearnerPrediction(learner, bh.task, features = c("lstat", "rm")) ``` # Support Vector Machine ```{r} learner = makeLearner("regr.ksvm") result = resample(learner, bh.task, rdesc, models = TRUE) getRRPredictions(result) ``` ```{r} getLearnerModel(result$models[[1]]) ``` ```{r} plotLearnerPrediction(learner, bh.task, features = c("lstat")) ``` ```{r} plotLearnerPrediction(learner, bh.task, features = c("lstat", "rm")) ``` # Regression Forests ```{r} learner = makeLearner("regr.randomForest") result = resample(learner, bh.task, rdesc, models = TRUE) getRRPredictions(result) ``` ```{r} getLearnerModel(result$models[[1]]) ``` ```{r} plotLearnerPrediction(learner, bh.task, features = c("lstat")) ``` ```{r} plotLearnerPrediction(learner, bh.task, features = c("lstat", "rm")) ``` ```{r} imp = getFeatureImportance(result$models[[1]]) imp sort(imp$res) ``` # More ```{r} getLearnerProperties(learner) getTaskDesc(bh.task) getTaskType(bh.task) ```