Chapter 4 - Predictive Analytics I: Data Mining

Question 1

What is Data Mining?

Answer 1

A term used to describe discovering knowledge from large amounts of data.

Question 2

How are companies dealing with data as it relates to understanding their customer?

Answer 2

They are analyzing the vast amount of data that they collect. Data mining helps the management of mission critical tasks with a high level of accuracy and timeliness.

Question 3

What are some reasons businesses have turned to Data Mining? (7)

Answer 3

1. More intense competition at the global scale
2. Untapped value hidden in large data sources
3. Consolidation and integration of database records
4. Consolidation of databases into a single location
5. Exponential increase in data processing & storage technologies
6. Significant reduction in cost of hardware and software
7. Movement toward demassification of business practices

Question 4

What is Genomic Data?

Answer 4

It combines genetics with statistical data analysis and computer science.

Question 5

What are Four Example Uses of Data Mining?

Answer 5

Used to detect and reduce fraudulent activities
Identify customer buying patterns and reclaim profitable customers
Identify trading rules from historical data
Aid in increased profitability using market-basket analysis

Question 6

What are the Seven (1-3) Characteristics and Objectives of Data Mining?

Answer 6

1. Data are cleansed and consolidated into a data warehouse
2. Data Mining environment is usually a client/server architecture or a Web-based IS architecture.
3. Sophisticated new tools help to remove information buried in corporate files or archival public records. Also explores the usefulness of soft data.

Question 7

What are the Seven (4-7) Characteristics and Objectives of Data Mining?

Answer 7

4. The miner is often the end user who obtains answers quickly
5. Striking it rich involves finding unexpected results and requires users to think creatively throughout the process
6. Data mining tools are readily combined with spreadsheets and other software development tools.
7. Due to the large amounts of data, it is sometimes necessary to use parallel processing for data mining.

Question 8

What are the Six (6) Multiple Disciplines associated with Data Mining?

Answer 8

1. Knowledge Extraction
2. Pattern Analysis
3. Data Archaeology
4. Information Harvesting
5. Pattern Searching
6. Data Dredging

Question 9

What are the Four (4) Major Types of Patterns Data Mining Seeks to Identify?

Answer 9

Association - Find the commonly co-occurring groupings
Predictions - Tell the nature of future occurrences of certain events based on what has happened in the past.
Clusters - Identify natural groupings of things based on their known characteristics.
Sequential Relationships - Discover time-ordered events

Question 10

What is the Main Difference between Data Mining and Statistics?

Answer 10

Statistics starts with a well-defined proposition and well-defined hypothesis whereas data mining starts with a loosely-defined discovery statement.

Question 11

What are the Fourteen (1-5) Industry Focuses where Data Mining can be Applied?

Answer 11

1. Customer Relationship Management (CRM)
2. Banking
3. Retailing and Logistics
4. Manufacturing and Production
5. Brokerage and Securities Trading

Question 12

What are the Fourteen (6-10) Industry Focuses where Data Mining can be Applied?

Answer 12

6. Insurance
7. Computer Hardware and Software
8. Government and Defense
9. Travel Industry (Airlines; Hotels; Rental Car Companies, etc.)
10. Healthcare

Question 13

What are the Fourteen (11-14) Industry Focuses where Data Mining can be Applied?

Answer 13

11. Medicine
12. Entertainment Industry
13. Homeland Security and Law Enforcement
14. Sports

Question 14

What does CRISP-DM Stand For?

Answer 14

Cross-Industry Standard Process for Data Mining

Question 15

What are the Six (1-3) Steps associated with CRISP-DM?

Answer 15

1. Business Understanding - The key element of any data mining study is to know what the study is for.
2. Data Understanding - Identify the relevant data from many available databases.
3. Data Preparation - Take data and prepare it for analysis by data mining methods.

Question 16

What are the Six (4-6) Steps associated with CRISP-DM?

Answer 16

4. Model Building - Modeling techniques are selected and applied to an already prepared dataset to address the specific business need.
5. Testing and Evaluation - Models are evaluated how they meet business objectives and to what extent.
6. Deployment - Exploration, organization, and presentation of data findings.

Question 17

What does SEMMA stand for? What are the Five (5) Steps to SEMMA?

Answer 17

Sample - Generate a representative sample of the data
Explore - Visualization and basic representation of the data
Modify - Select variables; transform variable representations
Model - Use statistical and machine learning models
Assess - Evaluate the accuracy and usefulness of the models

Question 18

What does KDD mean?

Answer 18

KDD is knowledge discovery in databases. The process of using data mining methods to find useful info and patterns in the data.

Question 19

What are the Five (5) Elements to KDD?

Answer 19

Data Selection
Data Preprocessing
Data Transformation
Data Mining
Interpretation/Evaluation

Question 20

What are the Main Differences between Data Mining Methods?

Answer 20

Classification learns the function between characteristics and their membership through a supervised learning process, whereas Clustering learns the relationship through an unsupervised learning process where only the input variables are presented to the algorithm.

Question 21

What is a Simple Split?

Answer 21

It is the process of splitting the data into two mutually exclusive subsets called the training set and the test set.

Question 22

What is the K-fold Cross-Validation?

Answer 22

Also called rotation estimation, the complete dataset is randomly split into k mutually exclusive subsets of approximately equal size.

Question 23

What are the Four (4) other Classification Assessment Methodologies?

Answer 23

Leave-one-out - Similar to K-Fold but every data point is used for testing once on as many models developed as there are data points.
Bootstrapping - Fixed number of instances from the original data are sampled for training, and the rest is used for testing
Jackknifing - Similar to Leave-One-Out, but one sample is left out at each iteration of the estimation process.
Area Under the ROC Curve - graphical assessment where true positive rate is plotted on the y-axis and false positive rate is plotted on the x-axis.

Question 24

What are the Seven (7) Classification Techniques?

Answer 24

1. Decision Tree Analysis
2. Statistical Analysis
3. Neural Networks
4. Case-Based Reasoning
5. Bayeslan Classifiers
6. Genetic Algorithms
7. Rough Sets

Question 25

Why are Ensemble Models for Predictive Analytics Effective?

Answer 25

Combining forecasts can improve accuracy and robustness of information outcomes, while reducing uncertainty and bias associated with individual models.

Question 26

What is a Decision Tree?

Answer 26

It recursively divides a training set until each division consists entirely or primarily of examples of one class. Each non-leaf node contains a split-point, which is a test on one or more attributes and determines how the data are to be split further.

Question 27

What is the Gini Index?

Answer 27

It is used in economics to measure the diversity of a population.

Question 28

What is Information Gain?

Answer 28

It is the splitting mechanism used in ID3, which is more widely known as the decision tree algorithm.

Question 29

What is Entropy?

Answer 29

It measures the extent of uncertainty or randomness in a data set

Question 30

What are Cluster Analysis Results typically Used For?

Answer 30

It is used for classifying items, events, or concepts into common groupings. It is commonly used in biology, medicine, genetics, social network analysis, anthropology, archaeology, astronomy, character recognition, and even MIS development.

Question 31

What is the Apriori algorithm?

Answer 31

It is the most commonly used algorithm to discover association rules. Given a set of itemsets, the algorithm attempts to find subsets that are common to at least a minimum number of the itemsets.