Predictive Analytics I: Data Mining: Process, Methods, and Algorithms

Download as pdf or txt
Download as pdf or txt
You are on page 1of 60

Chapter 4

Predictive Analytics I: Data Mining


Process, Methods, and Algorithms
Learning Objectives (1 of 2)

4.1 Define data mining as an enabling technology for


business analytics
4.2 Understand the objectives and benefits of data mining
4.3 Become familiar with the wide range of applications of
data mining
4.4 Learn the standardized data mining processes
4.5 Learn different methods and algorithms of data mining
Learning Objectives (2 of 2)

4.6 Build awareness of the existing data mining software


tools
4.7 Understand the privacy issues, pitfalls, and myths of
data mining
Opening Vignette (1 of 3)

Miami-Dade Police Department Is Using Predictive


Analytics to Foresee and Fight Crime
• Predictive analytics in law
enforcement
– Policing with less
– New thinking on cold cases
– The big picture starts small
– Success brings credibility
– Just for the facts
– Safer streets for smarter
cities
Opening Vignette (2 of 3)

Discussion Questions
1. Why do law enforcement agencies and departments like
Miami-Dade Police Department embrace advanced
analytics and data mining?
2. What are the top challenges for law enforcement
agencies and departments like Miami-Dade Police
Department? Can you think of other challenges (not
mentioned in this case) that can benefit from data
mining?
Opening Vignette (3 of 3)

3. What are the sources of data that law enforcement


agencies and departments like Miami-Dade Police
Department use for their predictive modeling and data
mining projects?
4. What type of analytics do law enforcement agencies
and departments like Miami-Dade Police Department
use to fight crime?
5. What does “the big picture starts small” mean in this
case? Explain.
What is data mining?
Data Mining Concepts and Definitions
Why Data Mining?
• More intense competition at the global scale.
• Recognition of the value in data sources.
• Availability of quality data on customers, vendors, transactions,
Web, etc.
• Consolidation and integration of data repositories into data
warehouses.
• The exponential increase in data processing and storage
capabilities; and decrease in cost.
• Movement toward conversion of information resources into
nonphysical form.
Definition of Data Mining

• The nontrivial process of identifying valid, novel,


potentially useful, and ultimately understandable patterns
in data stored in structured databases.
– Fayyad et al., (1996)

• Keywords in this definition: process, nontrivial, valid,


novel, potentially useful, understandable.
• Data mining: a misnomer?
• Other names: knowledge extraction, pattern analysis,
knowledge discovery, information harvesting, pattern
searching, data dredging,…
Figure 4.1 Data Mining is a Blend of Multiple
Disciplines
Application Case 4.1
Visa Is Enhancing the Customer Experience While
Reducing Fraud with Predictive Analytics and Data
Mining
Questions for Discussion
1. What challenges were Visa and the rest of the credit card
industry facing?
2. How did Visa improve customer service while also
improving retention of fraud?
3. What is in-memory analytics, and why was it necessary?
Data Mining Characteristics & Objectives

• Source of data for DM is often a consolidated data


warehouse (not always!).
• DM environment is usually a client-server or a Web-
based information systems architecture.
• Data is the most critical ingredient for DM which may
include soft/unstructured data.
• The miner is often an end user.
• Striking it rich requires creative thinking.
• Data mining tools’ capabilities and ease of use are
essential (Web, Parallel processing, etc.).
How Data Mining Works

• DM extract patterns from data


– Pattern? A mathematical (numeric and/or symbolic)
relationship among data items
• Types of patterns
– Association
– Prediction
– Cluster (segmentation)
– Sequential (or time series) relationships
Application Case 4.2
Dell Is Staying Agile and Effective with Analytics in the 21st
Century
Questions for Discussion
1. What was the challenge Dell was facing that led to their
analytics journey?
2. What solution did Dell develop and implement? What were
the results?
3. As an analytics company itself, Dell has used its service
offerings for its own business. Do you think it is easier or
harder for a company to taste its own medicine? Explain.
A Taxonomy for Data Mining
• A Simple Taxonomy for Data Mining Tasks, Methods, and
Algorithms Data Mining Tasks & Methods Data Mining Algorithms Learning Type

Prediction

Decision Trees, Neural Networks, Support


Classification Supervised
Vector Machines, kNN, Naïve Bayes, GA

Linear/Nonlinear Regression, ANN,


Regression Supervised
Regression Trees, SVM, kNN, GA

Autoregressive Methods, Averaging


Time Series Supervised
Methods, Exponential Smoothing, ARIMA

Association

Market-basket Apriory, OneR, ZeroR, Eclat, GA Unsupervised

Expectation Maximization, Apriory


Link analysis Unsupervised
Algorithm, Graph-based Matching

Apriory Algorithm, FP-Growth, Graph-


Sequence analysis Unsupervised
based Matching

Segmentation

Clustering K-means, Expectation Maximization (EM) Unsupervised

Outlier analysis K-means, Expectation Maximization (EM) Unsupervised


Other Data Mining Patterns/Tasks

• Time-series forecasting
– Part of the sequence or link analysis?
• Visualization
– Another data mining task?
• Data Mining versus Statistics
– Are they the same?
– What is the relationship between the two?
Data Mining Applications (1 of 4)

• Customer Relationship Management


– Maximize return on marketing campaigns
– Improve customer retention (churn analysis)
– Maximize customer value (cross-, up-selling)
– Identify and treat most valued customers
• Banking & Other Financial
– Automate the loan application process
– Detecting fraudulent transactions
– Maximize customer value (cross-, up-selling)
– Optimizing cash reserves with forecasting
Data Mining Applications (2 of 4)

• Retailing and Logistics


– Optimize inventory levels at different locations
– Improve the store layout and sales promotions
– Optimize logistics by predicting seasonal effects
– Minimize losses due to limited shelf life
• Manufacturing and Maintenance
– Predict/prevent machinery failures
– Identify anomalies in production systems to optimize
the use manufacturing capacity
– Discover novel patterns to improve product quality
Data Mining Applications (3 of 4)

• Brokerage and Securities Trading


– Predict changes on certain bond prices
– Forecast the direction of stock fluctuations
– Assess the effect of events on market movements
– Identify and prevent fraudulent activities in trading
• Insurance
– Forecast claim costs for better business planning
– Determine optimal rate plans
– Optimize marketing to specific customers
– Identify and prevent fraudulent claim activities
Data Mining Applications (4 of 4)

• Computer hardware and software


• Science and engineering
• Government and defense
• Homeland security and law enforcement
• Travel, entertainment, sports
• Healthcare and medicine
• Sports,… virtually everywhere…
Data Mining Process
• A manifestation of the best practices
• A systematic way to conduct DM projects
• Moving from Art to Science for DM project
• Everybody has a different version
• Most common standard processes:
– CRISP-DM (Cross-Industry Standard Process for Data
Mining)
– SEMMA (Sample, Explore, Modify, Model, and Assess)
– KDD (Knowledge Discovery in Databases)
Data Mining Process: CRISP-DM (1 of 2)

• Cross Industry Standard Process for Data Mining


• Proposed in 1990s by a European consortium
• Composed of six consecutive phases
– Step 1: Business Understanding Accounts for
– Step 2: Data Understanding ~85% of total
– Step 3: Data Preparation project time
– Step 4: Model Building
– Step 5: Testing and Evaluation
– Step 6: Deployment
Data Mining Process: CRISP-DM (2 of 2)
• Figure 4.3 The Six-Step CRISP-DM Data Mining Process →

• The process is highly repetitive and experimental (DM: art versus science?)

1 2
Business Data
Understanding Understanding

3
Data
Preparation
6
4
Deployment
Model
Data
Building

5
Testing and
Evaluation
Data Mining Process: SEMMA
• Figure 4.5 SEMMA Data Mining Process

• Developed by SAS Institute


Sample
(Generate a representative
sample of the data)

Assess Explore
(Evaluate the accuracy and (Visualization and basic
usefulness of the models) description of the data)

Feedback

Model Modify
(Use variety of statistical and (Select variables, transform
machine learning models ) variable representations)
Data Mining Process: KDD
• Figure 4.6 KDD (Knowledge Discovery in Databases) Process
Internalization

Data Mining
DEPLOYMENT CHART
Knowledge
“Actionable
PHASE 1 PHASE 2 PHASE 3 PHASE 4 PHASE 5

DEPT 1

DEPT 2

DEPT 3

DEPT 4

Data 1 2 3 4 5 Insight”
Transformation
Extracted
Patterns

Data
Cleaning Transformed
Data

Data
Selection Preprocessed
Data

Target
Data

Feedback

Sources for
Raw Data
Which Data Mining Process is the Best?
• Figure 4.7 Ranking of Data Mining Methodologies/Processes.

CRISP-DM

My own

SEMMA

KDD Process

My organization's

Domain-specific methodology

None

Other methodology (not domain specific)

0 10 20 30 40 50 60 70

Source: Used with permission from KDnuggets.com.


Application Case 4.4
Cancer DB 1 Cancer DB 2 Cancer DB n

Data Mining Helps in Cancer


Research Combined
Cancer DB

Questions for Discussion


Data Preprocessing
ü Cleaning
1. How can data mining be used ü Selecting
ü Transforming

for ultimately curing illnesses Partitioned data


(training &
Partitioned data
(training &

like cancer? testing) Partitioned data


(training &
testing)
testing)

Artificial Neural Logistic Random


2. What do you think are the Networks (ANN)
Training and
Regression (LR)
Training and
Forest (RF)
Training and

promises and major challenges calibrating the


model
calibrating the
model
calibrating the
model

for data miners in contributing Testing the


model
Testing the
model
Testing the
model
Assess
variable
importance

to medical and biological


research endeavors?
Tabulated Model Tabulated
Testing Results Relative Variable
(Accuracy, Sensitivity Importance
and Specificity) Results
Data Mining Methods: Classification

• Most frequently used DM method


• Part of the machine-learning family
• Employ supervised learning
• Learn from past data, classify new data
• The output variable is categorical (nominal or ordinal) in
nature
• Classification versus regression?
• Classification versus clustering?
Assessment Methods for Classification

• Predictive accuracy
– Hit rate
• Speed
– Model building versus predicting/usage speed
• Robustness
• Scalability
• Interpretability
– Transparency, explainability
Accuracy of Classification Models
• In classification problems, the primary source for accuracy
estimation is the confusion matrix

TP + TN
Accuracy = True/Observed Class
TP + TN + FP + FN
Positive Negative
TP
True PositiveRate =

Positive
True False
TP + FN

Predicted Class
Positive Positive
Count (TP) Count (FP)
TN
True NegativeRate =
TN + FP

Negative
False True
TP TP Negative Negative
Precision = Recall = Count (FN) Count (TN)
TP + FP TP + FN
Estimation Methodologies for
Classification: Single/Simple Split

• Simple split (or holdout or test sample estimation)


– Split the data into 2 mutually exclusive sets: training
(~70%) and testing (30%)

Model
Training Data Development
2/3

Trained Prediction
Preprocessed Classifier Accuracy
Data
1/3 Model TP FP
Assessment
Testing Data (scoring) FN TN

– For Neural Networks, the data is split into three sub-


sets (training [~60%], validation [~20%], testing [~20%])
Estimation Methodologies for Classification: k-
Fold Cross Validation (rotation estimation)

• Data is split into k mutual subsets and k number


training/testing experiments are conducted
• Figure 4.10 A Graphical Depiction of k-Fold Cross-
Validation
Additional Estimation Methodologies for
Classification

• Leave-one-out
– Similar to k-fold where k = number of samples
• Bootstrapping
– Random sampling with replacement
• Jackknifing
– Similar to leave-one-out
• Area Under the ROC Curve (AUC)
– ROC: Receiver Operating Characteristics (a term
borrowed from radar image processing)
Area Under the ROC Curve (AUC) (1 of 2)
• Works with binary classification
• Figure 4.11 A Sample ROC Curve
Area Under the ROC Curve (AUC) (2 of 2)

• Produces values from 0 1

to 1.0 0.9

0.8

• Random chance is 0.5 0.7


A

and perfect classification 0.6

0.5

is 1.0 0.4
Area Under the
ROC Curve
(AUC) A = 0.84
0.3

• Produces a good 0.2

assessment for skewed 0.1

class distributions too! 0


0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

False Alarms (1 - Specificity)


Classification Techniques
• Decision tree analysis
• Statistical analysis
• Neural networks
• Support vector machines
• Case-based reasoning
• Bayesian classifiers
• Genetic algorithms
• Rough sets
Decision Trees (1 of 2)
• Employs a divide-and-conquer method
• Recursively divides a training set until each division consists of
examples from one class:
1. Create a root node and assign all of the
A general training data to it.
algorithm
(steps) for 2. Select the best splitting attribute.
building a
3. Add a branch to the root node for each value of
decision tree
the split. Split the data into mutually exclusive
subsets along the lines of the specific split.
4. Repeat steps 2 and 3 for each and every leaf
node until the stopping criteria is reached.
Decision Trees (2 of 2)

• DT algorithms mainly differ on


1. Splitting criteria
▪ Which variable, what value, etc.
2. Stopping criteria
▪ When to stop building the tree
3. Pruning (generalization method)
▪ Pre-pruning versus post-pruning
• Most popular DT algorithms include
– ID3, C4.5, C5; CART; CHAID; M5
Ensemble Models for Predictive Analytics
• Combine multiple other models in the prediction process
• Produces more robust and reliable prediction models
• Figure 4.12 Graphical Illustration of a Heterogeneous Ensemble
Application Case 4.5

Influence Health Uses Advanced Predictive Analytics


to Focus on the Factors That Really Influence People’s
Healthcare Decisions
Questions for Discussion
1. What did Influence Health do?
2. What were the challenges, the proposed solutions, and
the obtained results?
3. How can data mining help companies in the healthcare
industry (in ways other than the ones mentioned in this
case)?
Cluster Analysis for Data Mining (1 of 4)

• Used for automatic identification of natural groupings of


things
• Part of the machine-learning family
• Employ unsupervised learning
• Learns the clusters of things from past data, then assigns
new instances
• There is not an output/target variable
• In marketing, it is also known as segmentation
Cluster Analysis for Data Mining (2 of 4)

• Clustering results may be used to


– Identify natural groupings of customers
– Identify rules for assigning new cases to classes for
targeting/diagnostic purposes
– Provide characterization, definition, labeling of
populations
– Decrease the size and complexity of problems for
other data mining methods
– Identify outliers in a specific domain (e.g., rare-event
detection)
Cluster Analysis for Data Mining (3 of 4)

• Analysis methods
– Statistical methods (including both hierarchical and
nonhierarchical), such as k-means, k-modes, and so
on.
– Neural networks (adaptive resonance theory [ART],
self-organizing map [SOM])
– Fuzzy logic (e.g., fuzzy c-means algorithm)
– Genetic algorithms
• How many clusters?
Cluster Analysis for Data Mining (4 of 4)

• k-Means Clustering Algorithm


– k : pre-determined number of clusters
– Algorithm (Step 0: determine value of k)
Step 1: Randomly generate k random points as initial
cluster centers.
Step 2: Assign each point to the nearest cluster center.
Step 3: Re-compute the new cluster centers.
Repetition step: Repeat steps 3 and 4 until some
convergence criterion is met (usually that the
assignment of points to clusters becomes stable).
Cluster Analysis for Data Mining - k-Means
Clustering Algorithm

• Figure 4.13 A Graphical Illustration of the Steps in the k-Means


Algorithm

Step 1 Step 2 Step 3


Association Rule Mining (1 of 6)
• A very popular DM method in business
• Finds interesting relationships (affinities) between variables
(items or events)
• Part of machine learning family
• Employs unsupervised learning
• There is no output variable
• Also known as market basket analysis
• Often used as an example to describe DM to ordinary people,
such as the famous “relationship between diapers and beers!”
Association Rule Mining (2 of 6)
• Input: the simple point-of-sale transaction data
• Output: Most frequent affinities among items
• Example: according to the transaction data…
“Customer who bought a lap-top computer and a virus
protection software, also bought extended service plan 70
percent of the time.”
• How do you use such a pattern/knowledge?
– Put the items next to each other
– Promote the items as a package
– Place items far apart from each other!
Association Rule Mining (3 of 6)

• A representative application of association rule mining


includes
– In business: cross-marketing, cross-selling, store
design, catalog design, e-commerce site design,
optimization of online advertising, product pricing, and
sales/promotion configuration
– In medicine: relationships between symptoms and
illnesses; diagnosis and patient characteristics and
treatments (to be used in medical DSS); and genes
and their functions (to be used in genomics projects)
–…
Association Rule Mining (4 of 6)

• Are all association rules interesting and useful?


A Generic Rule: X ⇒ Y [S%, C% ]

X, Y: products and/or services


X: Left-hand-side (LHS)
Y: Right-hand-side (RHS)
S: Support: how often X and Y go together
C: Confidence: how often Y go together with the X

Example: {Laptop Computer, Antivirus Software} ⇒


{Extended Service Plan} [30%, 70%]
Association Rule Mining (5 of 6)

• Several algorithms are developed for discovering


(identifying) association rules
– Apriori
– Eclat
– FP-Growth
– + Derivatives and hybrids of the three
• The algorithms help identify the frequent itemsets, which
are then converted to association rules
Association Rule Mining (6 of 6)
• Apriori Algorithm
– Finds subsets that are common to at least a minimum
number of the itemsets
– Uses a bottom-up approach
▪ frequent subsets are extended one item at a time
(the size of frequent subsets increases from one-
item subsets to two-item subsets, then three-item
subsets, and so on), and
▪ groups of candidates at each level are tested
against the data for minimum support
(see the figure)  --
Association Rule Mining Apriori Algorithm
• Figure 4.13 A Graphical Illustration of the Steps in the k-Means Algorithm

Raw Transaction Data One-item Itemsets Two-item Itemsets Three-item Itemsets

Transaction SKUs Itemset Itemset Itemset


Support Support Support
No (Item No) (SKUs) (SKUs) (SKUs)

1001234 1, 2, 3, 4 1 3 1, 2 3 1, 2, 4 3
1001235 2, 3, 4 2 6 1, 3 2 2, 3, 4 3
1001236 2, 3 3 4 1, 4 3
1001237 1, 2, 4 4 5 2, 3 4
1001238 1, 2, 3, 4 2, 4 5
1001239 2, 4 3, 4 3
Data Mining Software Tools
R 1,419
Python 1,325
SQL 1,029
• Commercial Excel
RapidMiner
972
944
Hadoop 641

– IBM SPSS Modeler Spark


Tableau
KNIME
536
521
624

(formerly Clementine) SciKit-Learn


Java
Anaconda
497
487
462
Hive 359

– SAS Enterprise Miner Mllib


Weka
Microsoft SQL Server
337
315
314
Unix shell/awk/gawk 301

– Statistica - Dell/Statsoft MATLAB


IBM SPSS Statistics
263
242
Dataiku 227
SAS base 225
– … many more IBM SPSS Modeler
SQL on Hadoop tools
222
211
C/C++ 210
Other free analytics/data mining tools 198
Other programming and data languages 197
• Free and/or Open Source H2O
Scala
193
180
SAS Enterprise Miner 162

– KNIME Microsoft Power BI


Hbase
QlikView
161
158
153 Legend:
Microsoft Azure Machine Learning 147 [Orange] Free/Open Source tools
– RapidMiner Other Hadoop/HDFS-based tools
Apache Pig
141
132
[Green] Commercial tools
[Blue] Hadoop/Big Data tools
IBM Watson 121
Rattle 103
– Weka Salford SPM/CART/RF/MARS/TreeNet
Gnu Octave
100
89
Orange 89

– R, … 0 200 400 600 800 1000 1200 1400 1600


Application Case 4.6 (1 of 5)

Data Mining Goes to Hollywood: Predicting Financial


Success of Movies

• Goal: Predicting financial success of Hollywood movies


before the start of their production process
• How: Use of advanced predictive analytics methods
• Results: promising
Application Case 4.6 (2 of 5)

A Typical Classification Problem

Dependent Variable

Class No. 1 2 3 4 5 6 7 8 9
Range >1 >1 > > 20 > 40 > 65 > 100 > 150 > 200
(in (Flop > 10 10 < 40 < 65 < 100 < 150 < 200 (Blockbuster)
$Millions) ) <
20
Application Case 4.6 (3 of 5)

Independent Variables

Number of
Independent Variable Possible Values
Values
MPAA Rating 5 G, PG, PG-13, R, NR
Competition 3 High, Medium, Low
Star value 3 High, Medium, Low
Genre Sci-Fi, Historic Epic Drama, Modern
Drama, Politically Related, Thriller,
10
Horror, Comedy, Cartoon, Action,
Documentary
Special effects 3 High, Medium, Low
Sequel 2 Yes, No
Number of screens 1 Positive integer
Application Case 4.6 (4 of 5)
The DM Process Map in IBM SPSS Modeler

Model
Development
process

Model
Assessment
process
Application Case 4.6 (5 of 5)

*Training set 1998 – 2005 movies; Test set : 2006 Movies


Table 4.6 Data Mining Myths

Myth Reality
Data mining provides instant, crystal-ball-like Data mining is a multistep process that requires
predictions. deliberate, proactive design and use.

Data mining is not yet viable for mainstream The current state of the art is ready to go for
business applications. almost any business type and/or size.

Data mining requires a separate, dedicated Because of the advances in database technology,
database. a dedicated database is not required.

Only those with advanced degrees can do data Newer Web-based tools enable managers of all
mining. educational levels to do data mining.

Data mining is only for large firms that have lots of If the data accurately reflect the business or its
customer data. customers, any company can use data mining.
Data Mining Mistakes

1. Selecting the wrong problem for data mining


2. Ignoring what your sponsor thinks data mining is and
what it really can/cannot do
3. Beginning without the end in mind
4. Not leaving sufficient time for data acquisition, selection,
and preparation
5. Looking only at aggregated results and not at individual
records/predictions
6. … 10 more mistakes… in your book

You might also like