Determining what factors explain alcohol consumption among students.
What factors explain excessive alcohol consumption among students? The record for the task sheet
comes from a survey of students who attended mathematics and Portuguese courses and contains
many interesting details about their sociodemographics, life circumstances and learning success.
The ordinal scaled variables Dalc and Walc give information about the alcohol consumption of the
students on weekdays and weekends.
Create a binary target variable alc_prob as follows:
- Calculate the Gini index for the target variable
alc_proband the Gini index for each variable with respect toalc_prob. Determine the 5 variables with the highest Gini Gain. - Learn 2 different decision trees with alc_prob as target variable. For the first tree, nodes should be further partitioned until the class distribution of all resulting leaf nodes is pure. For the second tree, nodes with a cardinality of less than 20 instances should not be further partitioned. Determine the quality of the trees by calculating sensitivity (True Positive Rate) and specificity (True Negative Rate) for a 70%:30% split in training and test sets. Display the decision trees graphically and discuss the differences in quality measures.
- Use
randomForest::randomForest()to create a random forest with 200 trees. As candidates for a split within a tree a random sample of 5 variables should be drawn. Calculate Accuracy, Sensitivity and Specificity for the Out-of-the-Bag instances and show the most important variables (?importance).
