R

Question 1

How do you start with data

Answer 1

Data Description, Data Segmentation, Data Correlation, Data Classification

Question 2

How do you summarise categorical variables

Answer 2

By Counts or By Percentage

Question 3

What two types of categorical variables exist and give an example for each

Answer 3

Ordered (low/medium/high)

Unordered (male/female)

Question 4

What is the motivation behind data description

Answer 4

To find the shape of the data set, and to summarise the columns and relationships. Find out what the column headers are, and what type of variables are within the columns

Question 5

Describe what data segmentation means

Answer 5

To segment data into different subgroups as there are too many subjects in the data

Question 6

What is the ABC method of analysis

Answer 6

A: 80%, B: 15%, C: 5%%

Question 7

How do you visualise data after you segment it

Answer 7

Pareto Plot, Pie Chart, Bar Graph

Question 8

What is data correlation

Answer 8

Finding out the relationships between the variables

Question 9

Is more “scatter” around the line indicative of higher or lower correlation?

Answer 9

Lower. A perfect positive correlation would see all dots on the line

Question 10

plot(creditdata[,c(5,9:14)]). Explain this line of code

Answer 10

plot the credit data graphs, all the rows, with columns 5, 9 to 14

Question 11

What are decision trees?

Answer 11

They help you make decisions, mainly for categorical responses

Question 12

What is the purpose of decision trees

Answer 12

They help classify records by assigning it to a likely class or category. Can also calculate probability of event.

Question 13

Are decision trees non linear or linear

Answer 13

Non linear

Question 14

What algorithm do decision trees use

Answer 14

Recursive Greedy algorithm. you start off with an empty tree. decide which feature to split on. if nothing more to split, then you can make prediction. if not, think about when to stop.

Question 15

how to decide which feature to split

Answer 15

see which one has the lowest classification error (number of mistakes/total)

Question 16

How to decide when to stop

Answer 16

all data agrees on one condition, or you have no more features to split on

Question 17

how do you split continuous variables?

Answer 17

use midpoints of data, use thresholds with lowest classification error

Question 18

how to code a decision tree in R

Answer 18

install package tree
library(tree)
credit = tree(...)
summary(credit)
plot(credit)
text(credit, pretty = 0)

Question 19

how to summarise continuous data

Answer 19

use mean and standard deviations. most observations should lie within 2 standard deviations of the mean.

Question 20

what does 95% confidence interval mean?

Answer 20

it means based on normal distribution, that the data is 1.96 standard deviations above and below the mean

Question 21

what would be useful to summarise categorical data?

Answer 21

pareto plots,percentages, bar charts, pie charts, abc classifications