Question $1.$

According to the authors, what is the role of data mining in healthcare?

Question $2.$

What unique challenges are presented by data in healthcare? You may also want to take a look at

$$Future Directions$$ section at the end of the paper.

Question $3.$

What is the CRISP$-$DM methodology in data mining and how is it used in healthcare domain? See

$$Healthcare Data Mining Applications$$ section.

$1$

Question $4.$

On Figure $1$ of Page $5,$ we are shown the model that classifies whether a certain patient is diabetic or not. According to the authors, what is the advantage of using a decision tree model for

healthcare domain?

Note. The Figure $1$ has a typo; the root node should have $$N$$ and $$D$$ where D should

represent $261$ patients that are diabetic.

Question $5.$

According to the authors, what is the attribute that is considered the most important in classifying

whether a patient is diabetic or not? What is the second most important attribute? Can you guess

how these are inferred from Figure $1?$

Hint: Focus on the second column of Page $6$ and see how the authors are able to find which

nodes of the decision tree are important.

Question $6.$

According to the authors, what are the limitations of data mining applications in healthcare domain? In particular, comment on

$$ What role does a data warehouse provide in data mining for healthcare?

$$ the type of data problems present for healthcare.

$$ limitations in the predictive modeling process for healthcare data.

$$ importance of domain experts

$$ scope of investment needed for applications

Question $7.$

After reading the paper, you propose to apply hierarchical clustering on the same dataset. The list

of attributes in the dataset include seven variables of particular interest: gender, age, body mass

index $($BMI$),$ waisthip ratio $($WHR$),$ smoking status, the number of times a patient exercises per

week, and onset of diabetes.

$$ Explain why imputation is crucial for healthcare data. Why do you think using imputation

can be problematic no matter which method is chosen? Hint: You may want to use your

answers for Question $2$ to motivate your answer for this part.

$$ Explain why scaling the data is necessary before applying any clustering algorithm.

$$ Explain how clustering obtained from a hierarchical clustering and clusterings obtained from

a k$-$means clustering can be qualitatively different.