Client Assignment Order: Please Bid for ref code 1575 Due Date 2014.09.27 3pm

MIS772 – Predictive Analytics

========================================= Assignment 3

========================================= MIS772 Predictive Analytics – Trimester II, 2014 Assignment 3 – Customer Profiling and Market Basket Analysis Release Date: 20th September 2014 Due Date: 12th October 2014 Weight: 33.3% Format of Submission: A report (electronic form) + electronic submission of project in CloudDeakin site (submission site and instructions will be provided closer to due date). PART A (40%) A segmentation based exploration of customers in the churn case study Carry out an exploratory analysis to try and understand who these customers are and whether they have any behavioral patterns and tendencies which could be made use of. Although this analysis will not be directly linked to the earlier predictive analytics exercise, the results may provide useful incites when making decisions based on the predictive analysis results. Add another copy of the churn_telecom data source to the churn case study diagram (you may use another new diagram). Go to the meta-data page and change variable roles to input (change the rejected ones to input), other than the ID roles. Add a cluster node and a segment profile node to the diagram. Link the data source to the cluster node and the cluster node to the profile node. Carry out the following clustering and profiling activities and report outcomes. (it is important to note that this is an exploration of the customer data set. Therefore there will be no correct or incorrect result. What is expected is a report on findings and where appropriate suggestions on possible value). Open the variable information page of the cluster node (from the properties list). Since we are planning to conduct a cluster analysis using a limited number of variables, change the ‘use’ column to ‘no’ for all variables.

1. Carry out a demographics based profiling. Change the use of variables age, gender and customer value to ‘default’ (we will take customer value as a demographics variable although this may not be so – insufficient demographics in the data). Run the cluster node and see results. What can you say about the demography based segments. Run the segment profile node and comment on the results.

2. Include some customer status based information in the analysis – eg: tenure on network, no. of active services, total profitability of subscription, no of emails, internet/fix line revenue etc (use at least 3 variables). Run the cluster node and the segment profile node and discuss the outcomes. (Do you see any understandable groupings/segments?)

School of Information and Business Analytics, Deakin University

MIS772 – Predictive Analytics

3. Remove the initial variables and carry out a cluster analysis of usage information with variables such as average number of outgoing calls, incoming calls, number of local/international calls etc (use at least 4 variables). Note- we can further carry out cross cluster analysis to link these segments to segments from previous analysis in 1 and 2 above but will not be required for this assignment.

Prepare a report (maximum 3 pages) based on the outcome of the first 3 steps. You may include screen shots of results and point out the variables of significance. The report must have a section discussing the potential value of these results when taking action based on a churn prediction and survival analysis. PART B – Discussion on the practical use of clustering, segmentation and profiling 20% Study the Roy Morgan value segments and Experian demographics segments and profiles in http://www.roymorgan.com/products/values-segments http://www.experian.com.au/consumer-segmentation/demographic-profiling- segmentation.html Using your knowledge on customer segmentation, clustering as well as module 2 lectures:

1. At what stages of a predictive analytics exercise will such information be useful?

2. How will you relate customer segments identified to Roy Morgan, Experian segments – how will this information be put to practical use?

Search for suitable references – you are expected to provide at least 2 references. Max 1 page. You may use the following as reference material: http://www.spatialinsights.com/catalog/downloads/products/80/How_Retailers_Can_Use _MOSAIC.pdf http://support.sas.com/resources/papers/proceedings12/286-2012.pdf PART C – Market Basket Analysis 40%

In order to plan innovative promotions to move items that are often purchased together, a store is interested in market basket analysis of items purchased from the Health and Beauty Aids Department and the Stationary Department. The store chose to conduct a market basket analysis of specific items purchased from these two departments. The TRANSACTIONS data set contains information about more than 400,000 transactions made over the past three months. The following products are represented in the data set: • bar soap • bows • candy bars • deodorant • greeting cards • magazines

School of Information and Business Analytics, Deakin University

MIS772 – Predictive Analytics

• markers • pain relievers • pencils • pens • perfume • photo processing • prescription medications • shampoo • toothbrushes • toothpaste • wrapping paper

There are four variables in the data set: Name Model

Role Measurement

Level Description

STORE Rejected Nominal Identification number of the store TRANSACTION ID Nominal Transaction identification number PRODUCT Target Nominal Product purchased QUANTITY Rejected Interval Quantity of this product purchased

a. Create a new diagram. Name the diagram Transactions.

b. Create a new data source using the data set ABA1.TRANSACTIONS.

c. Assign the variables STORE and QUANTITY the model role Rejected. These variables are not used in this analysis. Assign the ID model role to the variable TRANSACTION and the Target model role to the variable PRODUCT. Change the data source role to Transaction.

d. Add the TRANSACTIONS data set and an Association node to the diagram.

e. Change the setting for the Export Rule by ID property to Yes.

f. Leave the remaining default settings for the Association node and run the analysis.

g. Examine the results of the association analysis.

1. What is the highest lift value for the resulting rules

2. Which rules have this value?

3. What is the significance of the lift value of a rule – explain using an example from the case study.

4. What is the relationship between lift, support and confidence values – describe using an example.

5. Based on the association rules, briefly describe 3 example product bundles and promotions that you might suggest?

School of Information and Business Analytics, Deakin University