A Closer Look at Machine Learning

David Chappell
PRINCIPAL, CHAPPELL & ASSOCIATES

@DChappellAssoc www.davidchappell.com
Machine Training data
Learning Supervised and unsupervised learning
Concepts
Classifying machine learning problems
and algorithms
Training a model
Testing a model
Using a model
 Terminology

Training Supervised Unsupervised

Data Learning Learning

The prepared data The value you want to The value you want to
used to create a predict is in the predict is not in the
model training data training data
Creating a model is The data is labeled The data is unlabeled
called training a model

Most common
 Data Preprocessing with Supervised Learning
Features Target Value

Data Source 1 1) Read raw

data

Data 2) Create
Preprocessing training
Module(s) data

Data Source 2
...

Training Data
100011010011
110111110110

Data Source N
Available Data Preprocessing Modules
Categorizing Machine Learning Problems:
Regression

For supervised learning

Example question:
o How many units of this product
will we sell next month?
Categorizing Machine Learning Problems:
Classification

Class 1 Class 2 For supervised learning

Example question:
o Is this credit card transaction
fraudulent?

Can be more than

two classes
Categorizing Machine Learning Problems:
Clustering

Cluster 2
For unsupervised learning
Example question:
Cluster 1 Cluster 3
o What are our customer
segments?
 Styles of Machine Learning Algorithms:
Examples

P(A) P(B|A)
P(A|B) =
P(B)

Decision tree Neural network Bayesian K-means

 Training a Model with Supervised Learning
1) Choose features
Target
Feature 1 Feature 3 Feature 6 Value

Chosen
Candidate
Learning
Model
2) Input training data Algorithm 3) Generate
(75% of all data for candidate
features 1, 3, and 6) model

Training Data
 Testing a Model with Supervised Learning

Target
Feature 1 Feature 3 Feature 6 Value

Candidate
Model
1) Input test data 2) Generate
(remaining 25% of target values
all data for features from test data
1, 3, and 6)

Training Data
3) Compare target values
generated from test data with
actual target values
 Improving a Model: Some Options
1) Choose different features

Feature 1 Feature 2 Feature 5 3) Modify learning algorithm

parameters or choose a
different algorithm

Chosen
Learning Candidate
Algorithm Model

2) Add more (or newer) example data

Available Learning Algorithms
 Using a Model

Model

1) Call model with 2) Get back value

values for features predicted by the
the model requires model using those
features
Application
Summary Machine learning has come of age
Machine learning isn’t hard to understand
- Although it can be hard to do well

Machine learning can probably help your

organization