Pensum

Question 1

What are the two main machine learning techniques for data mining?

Answer 1

Supervised machine learning and unsupervised machine learning

Question 2

What are examples of supervised machine learning?

Answer 2

Decision trees, linear classifiers, linear regression

Question 3

What are examples of unsupervised machine learning?

Answer 3

Clustering

Question 4

What are characteristics with unsupervised machine learning?

Answer 4

No specific target value for unsupervised methods. System is just looking for pattern in the data but not acting like “a teacher”. Data can be grouped very nicely into a small number of categories. We just have to look for the result.

Question 5

What are the goal with clustering?

Answer 5

Goal is to group together similar instances using some metric of similarity - so create groupings where the members of a given group are similar to each other. For example group similar customers together and design different campaigns.

Question 6

What are characteristics with clustering?

Answer 6

It is light classification but the groupings are not predefined. More open ended than classification and regression. Could find a way to group similar customers together. May or may not relate to the churn question.

Question 7

What are the fundamental goal of data mining techniques?

Answer 7

Exploration to find patterns in dataset.

Question 8

What is similarity matching?

Answer 8

Instances are compared based on their attributes to determine how similar they are. Amazon - find books that are similar to a book you have read. The most similar will be a book with all three attributes (if there were three in the one you already read).

Question 9

When do we use similarity matching?

Answer 9

The general idea of similarity matching placeable in many different forms of data mining including classification, regression, and clustering

Question 10

What is important with similarity matching?

Answer 10

Important to have information about the relevant attributes. And information about which one attributes is most important.

Question 11

What is regression?

Answer 11

Numerical value. Related to classification but there is a difference. Classification predicts wether there is going to happen something. Regression predicts how much.

Question 12

Give an example of when we would use regression.

Answer 12

How much will a customer spend? that will be solved with regression.

Question 13

What does supervised machine learning do in general?

Answer 13

A target value specified for each instance. Examining instances one by one. We can simply compute how often the system makes the right choice.

Question 14

What is classification?

Answer 14

Classification involves defining a small number of classes and then trying to predict for each instance, which class they belong to. In churn example classification is a natural one - one for will churn and one for will not churn
Each instance is labelled with a target value indicating what class it belongs to.

Question 15

What is data preparation?

Answer 15

Data preparation is about constructing a dataset from one or more data sources to be used for exploration and modeling

Question 16

Why do we need data preparation?

Answer 16

It is a solid practice to start with an initial dataset to get familiar with the data, to discover first insights into the data and have a good understanding of any possible data quality issues. Data preparation is often a time consuming process and heavily prone to errors.

Question 17

What is a database?

Answer 17

Database collects, stores and manages information so users can retrieve, add, update or remove such information. It presents information in tables with rows and columns.

Question 18

What tools can you use to assess data?

Answer 18

Through accuracy, precision/ recall, testing/training, cross-validation

Question 19

What technique do you use when you want to find out how much a customer wants to use a service?

Answer 19

Regression

Question 20

What does classification do?

Answer 20

Predicts the class each individual belongs to

Question 21

What does regression do?

Answer 21

Estimates a numerical value for each individual

Question 22

What does clustering do?

Answer 22

Identifies similar individuals based on data known about them

Question 23

Can we find groups of customers who are likely to cancel the service when the contract expires?

Answer 23

This is a problem for supervised learning

Question 24

How can a CSV data file look like?

Answer 24

sunny,short,boring,no

Question 25

How do you calculate entropy? (which is a technique for information gain)

Answer 25

x

Question 26

A Linear Classifier is a Parameterized Model -- the Parameters are what is learned in the training process. What are the parameters for a Linear Classifier?

Answer 26

The weights

Question 27

A good way to recognise overfitting is:

Answer 27

Compare accuracy on holdout data with accuracy on training data

Question 28

kNN is a data mining technique that can be used for?

Answer 28

Classification and regression

Question 29

What are the most widely used techniques in data mining?

Answer 29

Classification, regression and clustering.

Question 30

What does the data mining technique co-occurrences and associations/market-basket analysis do?

Answer 30

Finding items that go together. For example, by analyzing market basket data, you might find that customers who bought a pork sandwich also bought a water. Learning these associations can be very useful.

Question 31

What is the core of data analytical thinking?

Answer 31

``` Data should be considered an asset Can help to structure business problems Applying data science to a well-structured problem vs exploratory data mining ```

Question 32

What is the aim of generalisation in data analytical thinking?

Answer 32

We want patterns that generalize to data we have not seen

Question 33

Mention four ways to extract knowledge from data

Answer 33

1. Identifying informative attributes 2. Fitting a numeric function model to data 3. Controlling complexity - generalization and overfitting 4. Calculating similarity between objects