Foundation

Question 1

How does Predictive Modelling work?

Answer 1

Predicts OUTCOME (target), based on set of INPUTS

Question 2

What are the 2 types of Prediction?

Answer 2

Classification, Estimation

Question 3

What criteria is needed to use “Classification”?

Answer 3

Target must be CATEGORICAL
Hint: “Class” in “Classification” => Categorical

Question 4

What criteria is needed to use “Estimation”?

Answer 4

Target must be CONTINOUS (numerical)
Hint: “Estimate” => Numbers hence, continuous

Question 5

What percentage should we split the data?

Answer 5

70% Training, 30% TESTING

Question 6

Using what node in SPSS Modeler can we split the data?

Answer 6

Partition node

Question 7

Why is it neccessary to SPLIT data?

Answer 7

Aim of predictive model: It should be trained to be accurate on UNSEEN data

What is UNSEEN data? It is the TRAINING data!

Question 8

Why we use “Seed” in IBM SPSS?

Answer 8

It helps to RANDOMLY select records (datarows) to be either training/testing data.

Question 9

Why remember the exact seed number?

Answer 9

To ensure that it does not randomly select a record to be training/testing data

This further ensures that the model’s result is consistent as the same records are chosen to be training and testing data respectively.

Question 10

What does CART stand for?

Answer 10

Classification And Regression Tree

Question 11

When to use Regression Tree?

Answer 11

When it is to estimate (the type of predict is to ESTIMATE numerical target)

Question 12

What are the 2 types of impurities to measure for Classification?

Answer 12

Gini Index, Entrophy

Question 13

What are the impurity measure for Regression?

Answer 13

Sum of Squared Error (SSE)

Question 14

Is lower or higher GINI better?

Answer 14

Lower! Because Gini shows impurity. Gini = 1-purity.

Question 15

Why is lower GINI more desirable?

Answer 15

The nodes are more homogenous = better prediction = better model

Question 16

State the 4 kind of nodes in a decision tree

Answer 16

Parent node, Child Node, Root node, Leaf node

Question 17

What is the main difference between a Confusion Matrix vs. Analysis Node that gives accuracy (correct/wrong)?

Answer 17

Confusion Matrix can give HIT rate by True positive, false negative, true false, false positive. Provides a more in depth detail.

Conversely, Analysis node only states Correct/ Wrong.

Question 18

Why is Confusion Matrix useful?

Answer 18

By analysing the True and Negative Positive/False, we can determine the severity (depending on context) to it’s hit rate. E.g) in a medical field, we would want MORE “False Positive” than “False Negative.

Why? Consequence is severe! So hint is to see the context’s consequences.

Question 19

Formula for False Positive Hit Rate

Answer 19

FP / FP + TN

Question 20

Formula for False Negative Hit Rate

Answer 20

FN / FN + TP

Question 21

What are the thing we have to take EXTRA note in decision trees?

Answer 21

Overfitting

Question 22

State the 2 signs of overfitting

Answer 22

1. Training accuracy > Testing Accuracy %
2. Leaf node only has 1 sample -> Gini = 0 -> 100% Pure (impossible because it is TOO specialized in training data)

Question 23

How can we prevent Overfitting?

Answer 23

1. Set rules to stop growth in IBM (“Overfitting Prevention”)
2. Prune tree until it does not overfit (see the best condition for MINIMUM ERROR RATE!!%)