[1]

Case Study Report

On
IBM SPSS Modeler for Data Mining
By

Shaik Rafi

1602-18-733-049

M.Shiva Ashwardh

1602-18-733-050

Submitting to

Ms.Sunitha

ASSISTANT PROFESSOR

Department of Computer Science & Engineering

Vasavi College of Engineering.

(Affiliated to Osmania University)

Ibrahimbagh, Hyderabad-31

2021
 [2]

TABLE OF CONTENTS
1 Abstract……………………………………………………...3

2 Introduction…………………………………………………..3

2.1 Overview…………………………………………………….3

3 Accessing all types of data ……………………..……………..5

4 Broadening your analytics reach with a range of techniques......7

5 Accommodating your needs with flexible deployment……......9

6 Execution and scheduling ………………………………….….10

7 Conclusion………………………………………………….….12
 [3]

1. Abstract:
The IBM SPSS Modeler is a data mining, modeling and reporting tool. It provides a nice
GUI to carry out all the data mining tasks in form of Nodes and Stream Flows.Nodes are
the icons or shapes that represent individual operations on the data. The nodes are linked
together in a stream to represent the flow of data through each operation i.e. A set of
actions (reading in, preprocessing, classification/association rule mining/clustering,
reporting, etc.) on some input data is called a stream.

2. Introduction
In a business environment, the main objective of analytics is to improve a business
outcome. These outcomes can include:
• Increasing revenue by reducing customer attrition
• Increasing cross sell rates with a call center
• Decreasing costs by identifying fraudulent claims before payment
• Servicing a component in a production line to minimise downtime.

2.1 Overview:

➢ When you apply analytics to improve a decision, the result is likely to be a better
outcome.
➢ Data mining is the process of using analytical techniques to uncover patterns in
data. Descriptive analysis, predictive modeling, text analytics, geospatial
analytics, entity analytics, decision management and optimisation are used to
identify patterns and deploy predictive models into operational systems. Systems
and people can use these patterns and models to derive insights that enable them
 [4]

to consistently make the right decision at the point of impact. Outcomes are
maximised based on the predictive intelligence hidden in data of growing size and
complexity.
➢ IBM® SPSS Modeler is a powerful predictive analytics platform that is designed
to bring predictive intelligence to decisions made by individuals, groups, systems
and your enterprise. SPSS Modeler scales from desktop deployments to
integration with operational systems to provide you with a range of advanced
algorithms and techniques. Applying these techniques to decisions can result in
rapid ROI (return on investment) and can enable organisations to proactively and
repeatedly reduce costs while increasing productivity.
➢ SPSS Modeler is available in four editions to meet virtually any analytical needs:
IBM SPSS Modeler Professional:
Uncover hidden patterns in structured data. SPSS Modeler Professional
provides advanced algorithms, data manipulation and automated modeling
and preparation techniques to build predictive models that can help you
deliver better business outcomes.
IBM SPSS Modeler Premium:
Unlock insights from almost any type of data using a range of advanced
algorithms and capabilities, such as text analytics, entity analytics and
social network analysis, along with automated modeling and preparation
techniques.
IBM SPSS Modeler Gold:
Build and deploy predictive models directly into your business processes
with decision management capabilities so people and systems can make
the right decisions every time. With SPSS Modeler Gold on Cloud, these
capabilities are available as a web-based monthly subscription service.
IBM SPSS Predictive Analytics Enterprise:
Optimise your decision-making at the point of impact with a single,
multifaceted predictive analytics solution that includes SPSS Modeler.
With each edition of SPSS Modeler, you can:
• Access all types of data
 [5]

• Broaden your analytics reach

• Accommodate your needs with flexible deployment.

3.Accessing all types of data:

Data is being generated at an exponential rate from a multitude of sources, thereby

fueling new information and untapped opportunities for those organisations able to
harness it and realise its value. This data is stored in various systems and formats so
bringing it together can be a challenge. The volume of data is so big that you cannot
analyse it manually nor can you look over tables in reports to find why something might
or might not happen. The analysis process presents yet another challenge because of a
scarcity of skilled analysts that can work with the data to extract its value.

With SPSS Modeler, you and your organisation can use the data you have available, both
spatial and non-spatial and extract value from it by discovering untapped opportunities
and new information. With new insights from your data, you can predict what is likely to
happen, become proactive and optimise outcomes, rather than reacting simply as your
current situation dictates.

SPSS Modeler enables you to use a range of analytical techniques to access data sources
such as flat files, databases, data warehouses, and Hadoop distributions. These statistical
techniques use historical data to make predictions about current conditions or future
events. Also included are capabilities for data access, data preparation, data modeling and
interactive visualisations. With automated procedures for preparation and modeling, it is
suitable for a wide range of analytics abilities.

The intuitive graphical interface of SPSS Modeler enables users to visualise each step of
the data mining process as part of a ‘stream.’ By interacting with these streams, analysts
and business users can collaborate, which adds business knowledge and domain expertise
 [6]

to the data mining process. Users can focus on discovering insights rather than on
technical tasks such as writing code. They can also pursue ‘train-of-thought’ analysis and
explore data more deeply, both of which can uncover additional relationships that make
sense to your organisation.

Data preparation and manipulation:

Preparing data for analysis is an important but time consuming step in analysis. SPSS
Modeler automates data preparation to ease the process and to help you make sure your
data is in the best format for analysis. The tasks automated include analysing data and
identifying fixes, screening out fields, deriving new attributes when appropriate, and
improving performance through intelligent screening techniques.

SPSS Modeler offers a variety of ways to manipulate and prepare data for analysis at the
record or field (or variable) level. Among the methods used to help make sure your data
is in the best format for the specific type of analysis that is being undertaken are:

• Record operations:

Select, Sample and Distinct nodes enable you to choose specific rows of data. You can
merge and append nodes to join data by adding columns or rows to a dataset. Aggregate
and Recency, Frequency, Monetary (RFM) Aggregate nodes summarise records to a
single row. A Balance node adjusts the proportions of records in imbalanced data and a
Sort node reorders based on value. The Space Time Box node creates geospatial and
time-based data for records.

• Field operations:

A Type node specifies metadata and properties of a dataset, and the Filter node discards
fields. The Derive node creates new fields and a Filler node can replace existing field
values. Data can be restructured with the Set to Flag, Restructure or Transpose nodes and
regrouped with the Reclassify or Binning nodes. To assist with modeling, the Partition
node can split the data and the History node and Time Intervals nodes can create
additional fields. The Field Reorder node defines the display ordering to make certain
fields easier to view.
 [7]

4. Broadening your analytics reach with a range of techniques:

Analytical techniques are continuing to evolve, providing analysts with a plethora of

options for tackling the problems in front of them. Additionally, as technology develops
and new types of data become available (such as location-based data from mobile phones
or cell towers), different questions and challenges arise about the best ways to exploit this
data. Innovative techniques are therefore necessary.

With SPSS Modeler, your analysts can solve their business problems with a single
platform that is designed to handle simple descriptive analysis, the most complex
optimisation problems and everything in between. SPSS Modeler features capabilities
that go beyond the standard analytic requirements of today’s analysts. A range of models,
automated modeling and data preparation, text analytics, entity analytics, social network
analysis and the ability to build models on parallel processes can all help you address the
most sophisticated problems.

Automated data modeling:

With the automated modeling features of SPSS Modeler, non-analysts can produce
accurate models quickly without specialised skills. In addition, advanced predictive
modeling capabilities enable professional analysts to create the most sophisticated of
streams.

Automated modeling enables you to compare multiple modeling approaches. By setting

specific options for each model type (or using the defaults), you can explore a multitude
of model combinations and options. The generated models are then ranked based on the
measure specified, saving the best for use in scoring or further analysis.

A range of models:

SPSS Modeler offers an array of modeling techniques, including all of the following
algorithms:
 [8]

• Classification algorithms. Make predictions or forecasts based on historical data with

techniques. Examples include decision trees, random trees, neural networks, logistic
regression, support vector machines, Cox regression, generalised linear mixed models
(GLMM) and more. Use automatic classification modeling for both binary and numeric
outcomes to streamline model creation or Self-Learning Response Modeling (SLRM) to
build a model that you can continually update or re-estimate without having to rebuild the
model.

• Segmentation algorithms. Group people or detect unusual patterns with techniques such
as automatic clustering, anomaly detection and clustering neural networks. Use automatic
classification to apply multiple algorithms with a single step and take the guesswork out
of selecting the right technique.

• Association algorithms. Discover associations, links or sequences with Apriori,

CARMA and sequential association.

• Time series and forecasting. Generate forecasts for one or more series over time with
statistical modeling techniques. Using temporal causal modeling, you can discover causal
relationships among a large number of series

• Extendibility with R and Python programming languages. Apply transformations, use

scripts to analyse, summarise or produce text and graphical output with R. With the
Custom Dialog Builder, you can share and reuse R and Python and Python code with
those who choose not to use programming for analysis.

• Monte Carlo simulation. Account for uncertainty in inputs to predictive models. Model
uncertain inputs based on historical data or with probability distributions to generate
simulated values, and then use them in the predictive model to generate an outcome. The
result is a distribution of outcomes that can provide answers to questions that are based
on realistically generated data .

• Entity analytics. Identify relationships and improve the coherence and consistency of
current data by resolving identity conflicts in the records themselves. Identifying these
 [9]

relationships can be vitally important in a number of fields, including customer

relationship management, fraud detection, anti-money laundering and security.

Geospatial analytics:

With SPSS Modeler, you can explore the relationship of data elements that can be tied to
a location and perform geographic spatial analysis of your data to reveal insights that
would not be visible in charts or tables. With spatial mining, you can easily mine
geospatial data using ESRI shape files. By analysing both non-spatial and spatial data,
overall model accuracy is improved and you are able to gain deeper insights into people
and events.

Add a new dimension to your analysis by discovering association rules among spatial and
non-spatial attributes. Using spatial temporal prediction, you can fit linear models for
measurements taken over locations in 2D space, enabling you to easily predict ‘hot’ areas
and how those areas will change over time. You can apply this technology to mine
geospatial data in fields such as crime pattern analysis, epidemic surveillance, building
management and branch performance analysis.

5. Accommodating your needs with flexible deployment:

The deployment of analytics in your organisation will depend on many environmental
factors. Such factors include the business problems that must be addressed, your choice
of operating systems and platforms, and the other technologies and data sources in your
infrastructure. Technology, and particularly software, should be flexible enough to
accommo date various permutations and still provide the expected performance and
results.

Decision management:

Decision management extends the predictive capabilities of SPSS Modeler to everyday

business processes to empower front-line employees and systems. It integrates predictive
 [10]

models, simple rules and scoring into your systems to automate, manage and optimise
high-volume decisions. It then recommends actions where and when people need them,
such as cross-selling on the phone with a customer, deciding the best routing for a claim,
using a utility to allocate bandwidth or presenting offers in a self-service kiosk.
Thousands of decisions can be made at the operational level in complete alignment with
your organisation’s goals and strategies.

With the decision management capability of SPSS Modeler:

• Predictive models can foresee the most likely outcomes and identify the factors driving
the outcome, such as the propensity of a customer to respond to a given offer or the risk
that a given claim is fraudulent.

• Business rules automate parameters that are determined by elements such as business
policies or legal and regulatory compliance. Basic rule support is provided directly in
SPSS Modeler. For more robust rules that scale to meet enterprise-wide requirements,
integration with IBM Operational Decision Management is also supported.

• Integrated scoring can provide almost instant recommend-ations to the right people and
systems so that resource-aware and strategically-aligned decisions can be made, no
matter the line of business.

6. Execution and scheduling:

SPSS Modeler includes capabilities that are designed to use automation to bring greater
consistency to your results. Greater consistency strengthens people’s confidence in
analytics because management can efficiently govern the business environments where
analytical processes take place. This governance helps ensure that all internal and
external procedural requirements are met.
With SPSS Modeler, your analysts can construct flexible, repeatable analytical processes
that can be operationalised, that is, initiated at the right time and integrated with other
enterprise processes. Predefined model management processes help models remain
relevant and accurate.
 [11]

In-database:
SPSS Modeler provides a number of capabilities to minimise data movement and push
analytics to the database, such as:
➢ SQL Pushback: With SPSS Modeler Server, moving data from large databases,
even in IBM System z and IBM PureSystems environments, is not required
because the analytics and mining can take place in the database. SQL Pushback
enables in-database data transformation and preparation without the need to write
any SQL or do any programming. The result is a significant improvement in
analytical performance.
➢ In-database scoring: Database-specific scoring adapters, which are available for
IBM SPSS Modeler with Scoring Adapter for zEnterprise, IBM DB2, IBM
PureData System for Analytics (powered by Netezza®) and Teradata solutions,
extend the number of SPSS Modeler algorithms that can be scored in database,
further reducing the need to extract the data before scoring.
➢ In-database mining:SPSS Modeler Server supports integration with the data
mining capabilities, modeling tools and database-native algorithms that are
available with IBM DB2 Analytics Accelerator (IDAA) on Z “ Hytap” (Hybrid
Transactional and Analytic Processing), PureDataSystem for Analytics, Oracle
Data Miner, Microsoft® Analysis Services and others. You can build, score, and
store models inside the database , all from the SPSS Modeler workbench.
Integration with IBM technology:
SPSS Modeler includes capabilities for exporting data to IBM Cognos Business
Intelligence and Cognos TM1 software. The results of analysis can be distributed for
reporting, monitoring and planning to key decision-makers who only need the results.
When further analysis is needed, SPSS Modeler can also access them as a data
source, which means the process can continue again, thereby feeding the results back
to wherever the questions began.
IBM SPSS Statistics provides the ability to carry out further statistical analysis and
data management to complement SPSS Modeler and its data mining abilities with a
dedicated section on the nodes.
 [12]

SPSS Modeler provides support for PureData System for Analytics to access specific
models from the SPSS Modeler Interface and leverage the hardware’s speed and
performance.

7. Conclusion:
SPSS Modeler is a predictive analytics platform that scales from desktop deployments to
integration in operational systems to bring predictive intelligence to decisions made by
individuals, groups, systems and the enterprise. Your organisation can use SPSS Modeler
to conduct analysis regardless of where the data is located, the size of the data, or whether
it is structured or unstructured. The scalable client-server architecture enables users to
access everything from flat files to big data environments. Analysis is pushed back to the
source for execution, minimising data movement and increasing performance.

With SPSS Modeler, all kinds of users can solve a variety of business problems. It offers
analytics techniques that range from descriptive analytics to advanced algorithms,
including automated modeling, text analytics, entity analytics, social network analysis,
decision management and optimisation. An intuitive interface is designed for a wide
range of users from the non-technical business user to the analytical professional. The
short learning curve for SPSS Modeler makes it appealing to the novice and advanced
user, so they can quickly uncover insights and realise real business results.