ISE 535 Data Mining - Homework 3: Segment and Brands
ISE 535 Data Mining
Homework 3
1 (40 pts) The file segment.csv has data from a survey to 300 customers and potential customers from a company offering a subscription service. It includes the age, gender, income, number of children, whether they own or rent their homes, and whether they currently subscribe to the service or not. It is of interest to know whether clients are just as likely to subscribe or not without regard to home ownership. a) Construct a single two-way crosstab table to count the number of individuals that are subscribers (or not) and are home owners (or not). b) Test if the proportion of subscribers is the same between home owners and home renters.
2 The file brands.csv has ratings from some perceptual attributes on a set of brands labeled as a,b,...,j. The data comes from a survey to 100 customers. The attributes are as follows.
| Perceptual attribute | Example |
|---|---|
| perform | Brand has strong performance |
| leader | Brand is a leader in the field |
| latest | Brand has the latest products |
| fun | Brand is fun |
| serious | Brand is serious |
| bargain | Brand products are a bargain |
| value | Brand products are a good value |
| trendy | Brand is trendy |
| rebuy | I would buy from Brand again |
a) (10 pts) Find the average rating of each brand on each attribute and store it in dataframe df1. Let column brand be the rownames of df1
rownames(df1) = df1$brand df1$brand = NULL
b) (10 pts) Display a heatmap using the average ratings from df1 using the following commands
library(gplots) library(RColorBrewer) heatmap.2(as.matrix(df1),col=brewer.pal(9, "GnBu"), trace="none", key=FALSE, dend="none",main="Brand attributes")
What brands are highly rated on attributes leader, and serious? c) (10 pts) Scale the brands data to find principal components (call prcomp1 the resulting object). How many principal components explain at least 80% of the variation in the customer’s brand ratings? Display a lineplot of the Cumulative PVEs. d) (10 pts) Construct a biplot from prcomp1 e) (10 pts) Find principal components from df1. Use prcomp2 = prcomp(df1,scale=T). f) (10 pts) Construct a biplot from prcomp2. This is called a perceptual map of the brands. It helps answer the question What is the average position of the brand on each attribute? What brands are highly rated on attributes leader, and serious? Which are highly rated on bargain, and value?
Submit your report (code and output) as a pdf file onto Blackboard (no screen captures).
