XLMiner

1
XLMiner a Data Mining Toolkit
www.xlminer.com

• QuantLink Solutions Pvt. Ltd.
• www.quantlink.com

2
XLMiner a quick tour

• Here is a short demo of XLMiner.
• Let us use a simple example
• a bank sends mailers to its customers offering a
  special deal on Personal Loans. In its previous
  campaign it got only about 9 positive response.
• Objective How to target customers for increased
  conversion rate.
• In other words the question to address is
• what profile indicates a high-potential customer

3
XLMiner a quick tour

• Past campaign data will be used to train the data
  mining model
• This is called supervised learning in DataMining
  terms
• Lets see how to build a model and use it for
  improving the response rate.

4
XLMiner Quick TourData description

• Our past campaign data has the following customer
  attributes
• Customer ID
• Customers Age
• Professional Experience
• Family Income
• Credit Card average annual spending
• Education Level
• appliances owned
• Did this customer accept past campaign offer
• The last variable is the known outcome of the
  past campaign. Our Data Mining model will use
  this for Supervised Learning.

5
XLMiner Quick TourA view of the data

• This is what the data looks like

The variable labeled as PersLoan is binary 0
means the customer was not interested in the
Personal Loan. 1 means the customer was
interested.
6
XLMiner Quick Tourthe Data Mining Process

Partition the data into Training
Validation Partitions
Fit the Model on Training Partition only
Obtain results see if they look good enough
Check if they are good for Validation data too!
Study the outputs for validation data
Try out several alternative models
Choose and deploy the best model
7
XLMiner Quick TourStart the analysis

• Lets get going with XLMiner.
• Notice that XLMiner is as easy to use as Excel!
• All we need to do is use the friendly menus. We
  follow just three simple steps to fit a model and
  see the outputs!

8
XLMiner Quick TourStep 1 Partition the data

• Well create two partitions by choosing the
  records randomly.
• The Training Partition will be used for fitting
  the model.
• The Validation partition will be used for
  checking if the model gives a good fit for
  another piece of known data.

9
XLMiner Quick TourPartitioned Data

• XLMiner creates a Partition Sheet that shows the
  data split into Two partitions.

Easy Hyperlinks on the Navigator facilitate
viewing of either partition
10
XLMiner Quick TourStep 2 Fitting the Model

• This is a Classification Problem where we want
  to predict customers as likely / not likely to
  take a Personal Loan.
• Lets use one of the available techniques
  Classification Tree.
• Later we can use other Classification techniques.

We select input (predictor) variables here
and the outcome variable here
11
XLMiner Quick TourStep 2 Fitting the Model

• The model fit guides us through easy wizard-like
  steps.
• In these steps we choose technique-specific
  parameters and the output options.
• In the end we click Finish to produce the
  results.

12
XLMiner Quick TourStep 3 Understanding the
Outputs

• The friendly Output Navigator lets us go over all
  the outputs.

The Summaries show us the classification error
percentages i.e. how well the model is
predicting
Many other diagnostic outputs are available
depending on options we choose.
Other outputs (like the Tree here) will tell us
the decision rules that the model is suggesting.
13
XLMiner Quick TourOutput 1 Validation Summary

• First we look at how well the model predicted
  for the Validation data set

In the Training data where we already knew the
outcome 156 will buy were predicted
correctly and 38 wrongly. 1801 Wont buy
were predicted correctly and merely 5 wrongly.
Here are the corresponding error
percentages. The errors are not very small but
could still indicate a workable model.
14
XLMiner Quick TourOutput 2 the decision rule

• Here is the Classification Tree that gives the
  easy-to-understand and implement Decision Rules

Cut-off points for different variables decide
whether to go Left or Right
0 not likely to buy 1 likely to buy
15
XLMiner Quick Tourthe decision rule in table form

• The same decision rule as shown visually can be
  converted into the table below. This is useful
  for implementing it in your information systems.

16
XLMiner Quick TourOutput 3 more details

• Each technique (Classification Tree in this case)
  has additional helpful outputs
• The example here shows the Prune Log how the
  percentage error reduced by pruning the tree

17
XLMiner Quick TourOutput 4 the Lift Chart

• Lift tells us how much better the model did
  compared to a random targeting of customers. This
  is one of the most important outputs.

With our Tree model we get a much superior
result. In less than 500 mailers sent to high
probability customers we would get nearly 170
successes!
If customers were targeted randomly we would
expect this outcome. For instance 1000 mailers
would probably yield less than 100 customers.
18
XLMiner Quick TourOutput 5 the Detailed report

• The Validation data is scored in detail as
  shown below. Scoring means using the fitted model
  to classify each record of the data.

Probability of success is computed for each
record. This is what helps XLMiner suggest
selective records (customers) to target.
Predicted values can be seen against the actuals
here.
19
XLMiner Quick TourTry several techniques!

• That was just one of the many techniques in
  XLMiner Classification Tree.
• A typical Data Mining exercise involves several
  alternative approaches on the same data. This can
  be either with different techniques or with
  different parameters or both.
• Comparing multiple approaches lets us assess
  which model to finally choose for implementation.

20
XLMiner Quick TourRich repertoire of techniques!

• XLMiner supports a comprehensive array of
  supervised learning procedures

• Multiple Linear Regression
• Logistic Regression
• Classification Regression Trees
• Neural Networks
• k Nearest Neighbors
• Naïve Bayes Classifier
• Discriminant Analysis

21
XLMiner Quick TourRich repertoire of techniques!

• ... and several other features in Unsupervised
  Learning Data Reduction and Exploration

• Principal Components Analysis
• k-means Clustering
• Hierarchical Clustering
• Self-organizing Maps (coming soon)
• Affinity Market Basket Analysis
• Here are some sample outputs from these methods

22
XLMiner Quick Toursample output - Dendrogram

• Hierarchical Clustering produces a dendrogram
  an excellent visual representation of Cluster
  formation.

Height of the bars is a measure of dissimilarity
in the clusters that are merging into one.
Smaller clusters agglomerate into bigger ones
with least possible loss of cohesiveness at each
stage.
23
XLMiner Quick Toursample output cluster
predictions

• Cluster Analysis has many powerful uses like
  Market Segmentation. We can view individual
  records predicted cluster membership.

24
XLMiner Quick Toursample output BoxPlots

• XLMiner supports powerful visualization. The
  example here shows BoxPlots of two variables.

Cluster 2 clearly shows higher Income Credit
Card spend than Cluster 1. This is an excellent
aid to characterizing the clusters
25
XLMiner Quick Toursample output Scatter Plots

• Matrix Scatterplots in XLMiner give a visual
  insight into relationship among variables.

26
XLMiner Quick Toursample output Association
Rules

• For Market Basket Analysis XLMiner produces
  easy-to-read Association Rules

Rules are explained in simple English!
Each rule tells us which offerings will go well
together
27
XLMiner Quick Tour and thats not all!

• XLMiner has handy utilities for Data handling
• Missing data treatment
• Transforming categorical data
• Binning continuous data
• Sampling from Databases
• Scoring to Databases

28
XLMiner Quick TourXLMiner gt Versatility!

• This was a quick demonstration of just a few
  things XLMiner can do.
• It can do lots more. It is comprehensive in
  coverage like the best DM products around.
• Get your free download for evaluation at
  www.xlminer.ncom

29
XLMiner Quick TourXLMiner gt Simplicity!

• Daryl Pregibon had said Data Mining is
  Statistics at Scale and Speed.
• Youll find that XLMiner is Statistics at Scale
  Speed and Simplicity!
• If you know to use Excel you already know
  XLMiner. You can get started in minutes.

30
XLMiner Quick TourXLMiner gt Great Value!

• Several comprehensive DM products are many times
  more expensive.
• For exploring how Data Mining will work for you
  XLMiner provides a great start!

31
XLMiner Quick TourWhat others say

• The American Statistician reviewed XLMiner along
  with other reputed products in the November 2003
  issue
• This is what it had to say
• An easy to use an excellent inexpensive add-on
  that greatly expands the capabilities of Excel.
• XLMiners documentation is remarkably good

32
XLMiner Quick TourMore Resources

• For your initiation into Data Mining
• Free evaluation download
• Online Courses at www.statistics.com
• Case Book in the making
• Technical references on product website

33
XLMiner Quick TourThank you for viewing this
Demo!

• www.xlminer.com