0% found this document useful (0 votes)
206 views32 pages

Chapter5-Technologies Enabling Information Insights and Decision

This chapter discusses technologies that enable business insights and decisions through data mining and analytics. It covers descriptive and predictive techniques for discovering patterns in data like decision trees, association analysis, and neural networks. The CRISP-DM process model is presented as a standard methodology for data mining projects. Case studies demonstrate how various organizations have successfully applied analytic techniques for applications like customer segmentation, risk analysis, and forecasting.
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
206 views32 pages

Chapter5-Technologies Enabling Information Insights and Decision

This chapter discusses technologies that enable business insights and decisions through data mining and analytics. It covers descriptive and predictive techniques for discovering patterns in data like decision trees, association analysis, and neural networks. The CRISP-DM process model is presented as a standard methodology for data mining projects. Case studies demonstrate how various organizations have successfully applied analytic techniques for applications like customer segmentation, risk analysis, and forecasting.
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 32

+

Chapter 5:Technologies
Enabling Insights and
Decisions
+
Introduction

 Data mining

 Business analytics

 Goal, design, techniques & implementation of data mining

 Decision trees or rule induction, as a knowledge-modeling tool

 Predictive techniques

 Real time decision support


+
Technologies to Create Insights: Using
Data Mining to Create New Explicit
Knowledge
 Business analytics

 Knowledge discovery in databases (KDD), or more commonly, data mining


(DM)

 Analytics comes with hypothesis testing

 Data mining is more the act of discovery that lacks a hypothesis

 DM applications have made noteworthy contributions to scientific discovery,


for example, in breast cancer diagnosis
+ Technologies to Create Insights: Using Data
Mining to Create New Explicit Knowledge

 Some of the factors driving BI are:


 Exploding data volumes
 Increasing decision complexity
 Need for quick reflexes
 Technological progress
+ BI Intelligence in Practice: Data Mining
Applications for e-Commerce

 One successful e-tailer implementation is the case of eBags, a web based


storefront of handbags, suitcases, wallets, and other similar products.

 Web pages garnering maximum purchases

 Utilization of customer information for customizing and maximizing profits

 Proflowers, an online resource that describes itself as ‘connecting consumers


with fresh-from-the-field flowers’

 Timely delivery of the product to maximizing customer satisfaction

 Improved the management


+
Other Successful DM Applications

 Banking

 Target Marketing

 Insurance

 Telecommunications

 Operations Management

 Retail Sales Forecasting

 Systems Diagnosis
+
The Business Analytics Process

 Business analytics

 Two significant ways to get new insights from the existing information:
 Discovery by using existing information
 Discovery by finding useful patterns in observations

 Cross-Industry Standard Process for Data Mining (CRISP-DM)


+
CRISP-DM Process Methodology
+
Steps for the Data Preparation

 Selection

 Construction and transformation of variables

 Data Integration

 Formatting
+
The Business Analytics Process

 DM process comprises the:


 Model building and validation
 Model evaluation and interpretation
 Deployment

 Customer Profile Exchange (CPEX)

 Data Mining Group (DMG)


+
The Data Mining Model:
Describing What Happened

 Describe what happened or to predict what will happen

 Descriptive techniques are used to look for patterns

 Descriptive techniques can be of two types:


 Affinity or association
 Clustering

 Market basket or association analysis can include the use of two techniques:
 Apriori association rule algorithm
 Generalized rule induction (GRI)
+
Association Techniques
+ BI in Practice: Harrah’s Customer Clusters
Lead to Higher Firm Profitability

 Use of technology for the improvement of customer satisfaction

 Company leadership embarked in the implementation of a customer loyalty


program

 Motivated by the economic slowdown caused by the dot bust of the late
1990s

 Innovative rewards program for improving sales, customer service and


marketing

 Included an enterprise-wide data warehouse that integrated customer data

 Invested in an award-wining CRM system used to analyze the customer


information for new patterns and insights
+
The Data Mining Model-Predicting
What Will Happen

 DM techniques used to predict what will happen are classified into three
categories:
 Statistical Data Mining methods
 Connectionist methods: Neural Networks
 Symbolic Data Mining methods :Rule induction
+
Statistical Data Mining Methods

 Find the correlations between the variables

 Curve fitting

 There are a number of curve fitting methods


 Least squares method
 Nonlinear correlation method
 Multivariate correlation techniques
 Inferential statistical techniques
+
Applicability of Statistical Techniques
+
Connectionist Methods: Neural
Networks
 Use of artificial neural networks (ANN) techniques

 A predictive technique (backward propagation network) or as a clustering


technique (Kohonen network)

 Most important feature of ANN is that they can ‘learn’

 Most popular neural network algorithm is the multilayered feed forward


neural network with the back propagation learning

 The process is continuously iterative

 Memory-based reasoning
+
Connectionist Methods: Neural
Networks

Inputs Outputs

Two-layer Neural Network


+
Applicability of Non-Inferential
Predictive Technique
+
Symbolic Data Mining Methods:
Rule Induction

 Decision Tree and Rule Induction Methods, also known as symbolic


techniques

 Used to infer the ‘rules’ that classify or partition the dataset

 Rule induction methods provide automated techniques

 The Classification and Regression Tree (CART) algorithm

 Chi-squared Automatic Interaction Detector (CHAID) and C5.0

 Root node-parent node-child node


+
Symbolic Data Mining Methods:
Rule Induction

 Classification Rule

 Traditional statistical prediction methods

 Decision Tree

 Variables

 Advantage-the results may be directly inspected to understand the variables


that can be effectively used to classify the data
+
Simple Decision Tree-(Source:
[SPSS, 2000])
+
Simple Decision Tree-(Source:
[SPSS, 2000])

 Depicts the outcome decision tree for a targeting marketing initiative

 Identifies ‘Age in Years’ as the most significant predictor for customer


purchases

 Node was further exploited using the ‘Income’ variable

 The decision points of the trees can intuitively be translated into software
conditional statements
+
Applicability of Decision Tree
Techniques
+
The Data Mining Model:
Understanding What Happened

 Evaluate the cost of the errors

 Shaded quadrant represents patients that were correctly predicted as being


healthy

 Shaded quadrant - 96 patients were diagnosed with presence of disease

 70.6% of the patients were correctly classified with the prediction algorithm
+
Classification Table Results [SPSS,
2000]
+ Case Study-Can knowledge discovery uncover the
factors that help predict the competitiveness of
nations?

 A nation’s competitiveness

 Productivity

 Competitiveness

 Country risk

 Competitiveness can be defined in one of the following ways:


 Cost-competitiveness
 Price-competitiveness
 Non-price competitiveness
+ Case Study-Can knowledge discovery uncover the
factors that help predict the competitiveness of
nations?

 WCY provides a competitiveness score for each country based on eight


major factors:
 Domestic economy
 Internationalization
 Government
 Finance
 Infrastructure
 Management
 Science and technology
 People

 Mathematical programming modeling


+ Predictors for Country Competitiveness According
to Different Data Mining Algorithms
+ Case Study-Can knowledge discovery uncover the
factors that help predict the competitiveness of
nations?

 The figure described country competitiveness using each of the four


methods:
 Stepwise regression (SWR)
 Weighted non-linear programming (WNLP)
 Neural networks (NN)
 Classification and regression trees (CART)
+
Effective Implementation of
Business Analytics
 Storage Law

 Data tombs or data stores

 Eight common mistakes that organizations seeking the deployment of DM


technologies must avoid:
 User expectations are too high
 Putting the right tools in the wrong hands
 Dishing up data that users need to figure out how to use
 Training users only at the beginning of the project
 Going for a quick win rather than planning for the long-haul
 The organization goes for the big bang
 Data roles and governance are not adequately addressed.
 The organization fails to demonstrate values.
+
Real-Time Decision Support:
Integrating Results With Action
 Real time decision support

 “interactive” as well as “invisible”

 Making analytics invisible requires the following challenges be addressed:


 Scaling analysis to large databases
 Scaling to high-dimensional data and models
 Automating the search
 Finding patterns and models understandable and interesting to user

 Vertically integrated decision support solutions

 Metadata

 Integration of existing environments

You might also like