ASSOCIATION RULE MINING

1. If {1, 2, 3}, {1, 2, 4}, {1, 3, 4}, {1, 2, 5}, and {3, 4, 5} are ALL the large 3-itemsets, list all of the large 2-itemsets out of the original data set - {1, 2, 3, 4, 5}.

2. Given the following (incomplete) data set about users of a web site and its document search and retrieval history, describe how you will apply association rule mining to the given data set by identifying the equivalents of items, item-grouping attribute (e.g., transaction) as well as the potential use of the patterns you obtain from the data set. Hint: You don't have to use all of the attributes in each answer

User ID	Session ID	Key words	Documents
U1	S1	K1 K2	D1
U1	S2	K2 K3 K4	D2
U1	S3	K2	D1
U1	S4	K3 K4 K5	D3
U1	S5	K4 K5	D2
U2	S6	K1 K2 K6	D1
U2	S7	K2 K6 K7	D2
U2	S8	K2 K6	D1
U2	S9	K8 K6 K2	D3
U2	S10	K2 K6 K9	D2
...	...	...	...

CLUSTERING

Given the distance function of the following data set, show the dendrogram generated by Bottom-up hierarchical clustering methods (using min cluster distance)

	1	2	3	4	5
1	0	3	9	9	5
2	3	0	8	7	6
3	9	8	0	2	2
4	9	7	2	0	1
5	5	6	2	1	0

I will give you the first two steps for "free":

Step 1. Find the minimal distance, which is 4 - 5, form the first cluster

Step 2. Re-build the table as following:

	1	2	3	4/5
1	0	3	9	7
2	3	0	8	6.5
3	9	8	0	2
4/5	7	6.5	2	0

CLASSIFICATION

1. Nave Bayes. Below is a data set about stolen cars. We have a YELLOW SUV DOMESTIC. Predict whether it is stolen using Nave Bayes. Show your process.

2. Consider the training dataset given below. In the dataset, "Purchase" is the label.

Customer	Age_group	Income_level	Education_level	Purchase?
1	0	1	0	Yes
2	0	0	0	No
3	1	1	1	Yes
4	1	0	1	No
5	0	1	1	No

Which attribute has the highest information gain, i.e., reduction in Gini Index? Justify your answer. Use Gini Index (1 - p₁² - p₂²)as measurement criteria

3. Two image recognition classifiers are being compared for their performances using confusion matrix:

a. which classifier has higher recall rate? Show your computation

b. if this is an app developed for self-driving cars to recognize bikes in the road so as to avoid accidents, which measurement metrics should we optimize? Precision or recall? Why?

4. Selecting Data Mining Tasks and Methods

The following describe the demographic, online browsing/shopping session and purchasing transaction information collected from a set of voluntary Internet users.

Table Name: Demographics
Household_id	num	Unique household identifier
hoh_most_education	num	Household Most Education
census_region	num	Census Region of Residence
household_size	num	Household Size
hoh_oldest_age	num	Household Eldest Age
household_income	num	Household Income
child_present	num	Child Present
racial_background	num	Racial Background
connection_speed	num	Connection Speed
country_of_origin	num	Country of Origin
Table Name: Session
Session_id Household_id date	num num num	Unique session identifier Household that signed on the session Date
time	num	Time
domain_id	num	Domain ID
domain_name	char	Domain Name
site_category_id	num	Site Category ID
duration	num	Duration
page_viewed	num	Page Viewed
ref_domain_name	char	Referring Domain Name
Table Name: Transaction
Trans_id Session_id prod_category_id	Num num num	Unique transaction identifier Session in which the transaction was completed Product Category ID
prod_totprice	num	Product Total Price
prod_name	char	Product Name
basket_tot	num	Basket Total

Answer the following questions:

a. Can you perform association rule analysis in this data set? Justify your answer and state any assumption you make about the attributes and/or attribute values if necessary.

b. eBay is interested in finding out what it is about an eBay customer that can help predict if the customer will buy large, moderate or small amounts (in dollars) of pre-owned products. Please recommend a data mining task that should be performed to find out what eBay is interested in and state the selection of all necessary attributes and their roles in the data mining task. Justify the recommendation and state any assumption you make about the attributes and/or attribute values if necessary. Please note that it is possible to find the insights in which eBay is interested via different data mining tasks with different choices of attributes respectively. You only need to specify one. As always, be creative and relevant at the same time.