In-class Quiz ISLP Chapter 8

Classification error rate of a region can not be defined as

• the fraction of training observations in that region that do not belong to the most common class
• the fraction of training observations in that region that belong to the most common class
• the fraction of test observations in that region that belong to the most common class
• None of the observations

Which of the following is NOT true about tree based methods?

• This method is about segmenting the predictor space into regions
• Prediction is based on mean or mode response value in a region
• Tree based methods are simple and useful for interpretation
• Tree based methods are one of the best supervised learning approach
• None of the above

Consider the following figure. What is the predicted salary of a player who played more than 4.5 years in Major League Baseball and scored more than 117.5 hits?

• approximately 6.74 million dollar
• approximately 6740 thousand dollar
• approximately 845 thousand dollar
• approximately 845 million dollar
• None of the above

Consider the following figure. Why splitting started with "Years" variable instead of "Hits" variable?

• "Years" has larger coefficient than "Hits" variable
• Starting the splitting with "Years" produces the lower RSS compared to "Hits"
• Starting the splitting with "Years" produces the higher RSS compared to "Hits"
• Starting the splitting with "Years" produces higher R^2 compared to "Hits"
• None of the above

Consider the residual sum of square (RSS) given in the mathematical expression given below. Tree based methods find the regions R1, R2 ......Rj that minimizes this RSS. The inner summation means

• Sum of squared distance of each observation from the mean value of a particular region

• Sum of squared distance of each region from the mean value of a particular region

• Sum of squared distance of each observation from the mean value of all regions

• None of the above

Consider the residual sum of square (RSS) given in the mathematical expression given below. Tree based methods find the regions R1, R2 ......Rj that minimizes this RSS. The outer summation means

• Sum of squared distance of each observation from the mean value of a particular region

• Sum of squared distance of each region from the mean value of a particular region

• Sum of squared distance of each observation from the mean value of corresponding regions

• None of the above

Recursive binary splitting is a greedy approach because

• best split at a particular step is made based on all possible information including future steps
• best split at a particular step is made based on all possible information in previous steps
• best split at a particular step is made based on information on that particular step
• None of the above

Classification tree is highly sensitive for tree growing. True or false?

• True
• False

Consider the following mathematical expression which defined Gini index: Which of the following statement is true about Gini index?

• Gini index is a measure of total variance across the available classes
• Gini index takes on a small value if most of the observations belongs to a single class
• Gini index takes on a high value if the value of node purity is high
• All of the above
• None of the above

Which of the following statement is NOT true?

• Gini index is used to evaluate the quality of a split
• Entropy is used to evaluate the quality of a split
• Classification error rate is preferable if prediction accuracy is the goal
• All of the above
• None of the above